List your company

Refusal mechanism ablation on JBB-Behaviors 100 harmful prompts (test)

100Refusal RateCAA
-4235077May 12, 2026

Evaluation results

MethodLinks
CAAModel=Qwen2.5-1.5B-Instruct, Steering Strength (alpha)=1.02026.051000.888
CAAModel=Qwen2.5-72B-Instruct, Steering Strength (alpha)=1.02026.05980.406
BaselineModel=Qwen2.5-1.5B-Instruct, Steering Strength (alpha)=0.02026.0597—
BaselineModel=Qwen2.5-3B-Instruct, Steering Strength (alpha)=0.02026.0592.9—
BaselineModel=Llama-3.1-8B-Instruct, Steering Strength (alpha)=0.02026.0588.9—
BaselineModel=Qwen2.5-7B-Instruct, Steering Strength (alpha)=0.02026.0582.8—
BaselineModel=Llama-3.1-70B-Instruct, Steering Strength (alpha)=0.02026.0572.7—
BaselineModel=Qwen2.5-72B-Instruct, Steering Strength (alpha)=0.02026.0565.7—
BaselineModel=Llama-3.2-3B-Instruct, Steering Strength (alpha)=0.02026.0557.6—
BaselineModel=Llama-3.2-1B-Instruct, Steering Strength (alpha)=0.02026.0543.4—
CAAModel=Llama-3.1-8B-Instruct, Steering Strength (alpha)=1.02026.0538.40.493
CNAModel=Qwen2.5-3B-Instruct, Steering Strength (alpha)=1.02026.0534.30.984
CNAModel=Llama-3.2-3B-Instruct, Steering Strength (alpha)=1.02026.0526.30.977
CNAModel=Qwen2.5-1.5B-Instruct, Steering Strength (alpha)=1.02026.0526.30.982
CNAModel=Llama-3.2-1B-Instruct, Steering Strength (alpha)=1.02026.0520.20.975
CNAModel=Qwen2.5-7B-Instruct, Steering Strength (alpha)=1.02026.0513.10.98
CNAModel=Llama-3.1-70B-Instruct, Steering Strength (alpha)=1.02026.0512.10.981
CNAModel=Llama-3.1-8B-Instruct, Steering Strength (alpha)=1.02026.055.10.969
CAAModel=Qwen2.5-7B-Instruct, Steering Strength (alpha)=1.02026.055.10.414
CNAModel=Qwen2.5-72B-Instruct, Steering Strength (alpha)=1.02026.055.10.983
CAAModel=Llama-3.2-1B-Instruct, Steering Strength (alpha)=1.02026.0500.554
CAAModel=Llama-3.2-3B-Instruct, Steering Strength (alpha)=1.02026.0500.431
CAAModel=Llama-3.1-70B-Instruct, Steering Strength (alpha)=1.02026.0500.569
CAAModel=Qwen2.5-3B-Instruct, Steering Strength (alpha)=1.02026.0500.844