layer,n_must_block,n_must_pass,caught,missed,wrongly_blocked,let_through,blocked_total,catch_rate,over_refusal_rate,refusal_precision,youden_j
layer 1 · input filter (instructor),10,10,4,6,1,9,5,0.4,0.1,0.8,0.3
layer 1 · PARANOID variant (the control experiment),10,10,10,0,10,0,20,1.0,1.0,0.5,0.0
layer 2 · scope judge (model:qwen2.5:3b-instruct) @ t>0.50,10,10,10,0,8,2,18,1.0,0.8,0.5556,0.2
layer 3 · output schema (instructor),5,5,5,0,0,5,5,1.0,0.0,1.0,1.0
THE CHAIN · first refusal wins,10,10,10,0,8,2,18,1.0,0.8,0.5556,0.2