List your company

Refusal Detection on JailbreakBench Refusal OOD (test)

100MC RecallACS Refusal Probe
95.8496.929899.08May 11, 2026

Evaluation results

MethodLinks
ACS Refusal ProbeTarget Model=Llama2026.0510083
ACS Refusal ProbeTarget Model=Qwen2026.0510080
ACS Refusal ProbeTarget Model=Mistral2026.0510070
ACS Refusal ProbeTarget Model=Phi2026.0510090
ACS Refusal ProbeTarget Model=LQMP (4-fam)2026.0510081
ACS Refusal ProbeTarget Model=LQMPG (5-fam)2026.059978
ACS Refusal ProbeTarget Model=Gemma2026.059668