List your company

Jailbreak Evaluation on JBB-Behaviors (ASR via GPT Judges, ACQ)

99.2ASR (GPT-4o)HMNS
38.67254.38670.185.814Apr 11, 2026

Evaluation results

MethodLinks
HMNSModel=Yi-1.5-72B-Chat2026.0499.295.61.8
HMNSModel=Mistral-7B-Instruct-v0.22026.0498.293.11.9
ArrAttackModel=Yi-1.5-72B-Chat2026.0494.690.37.1
HMNSModel=Qwen2.5-14B-Instruct2026.0494.588.41.9
ArrAttackModel=Mistral-7B-Instruct-v0.22026.0492.487.57.4
ArrAttackModel=Qwen2.5-14B-Instruct2026.0489.683.77.9
AutoDANModel=Yi-1.5-72B-Chat2026.047568.911.9
AutoDANModel=Mistral-7B-Instruct-v0.22026.0471.665.512.2
AutoDANModel=Qwen2.5-14B-Instruct2026.0466.159.413.2
FITDModel=Yi-1.5-72B-Chat2026.0447.241.515.1
FITDModel=Mistral-7B-Instruct-v0.22026.0443.237.615.8
FITDModel=Qwen2.5-14B-Instruct2026.044135.116.6