You need to agree to share your Hugging Face information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

WBench Evaluation Results

This directory uses the same metric JSON and report formats as the official WBench evaluation and contains results for 158 Navi cases. The aggregate report, report.json, is located in the root directory.

Directory Structure:

wbench/
├── README.md
├── icon.png
├── meta.json
├── report.json                    # Aggregate metric results
├── videos/
│   └── case_<id>.mp4              # 158 videos
└── evaluation/
    └── <metric>/
        └── case_<id>.json         # Per-case metric results

VLM Evaluation Result Paths

Per-case results for each metric are stored in the directories listed below. All paths are relative to the root of this package. Each result file is named case_.json and contains the score and scoring details for the corresponding case.

Scene Adherence (scene_adherence)
Number of cases: 158
Results directory: evaluation/scene_adherence/

Subject Adherence (subject_adherence)
Number of cases: 74
Results directory: evaluation/subject_adherence/

Causal and Physical Fidelity (causal_fidelity)
Number of cases: 50
Results directory: evaluation/causal_fidelity/

Aggregate report: report.json

Downloads last month
152