Jailbreak on StrongReject
96.1ASR (GPT-4o)HMNS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| HMNSModel=Yi-1.5-72B-Chat2026.04 | 96.1 | 92.3 | 2.1 | |
| HMNSModel=Mistral-7B-Instruct-v0.22026.04 | 93 | 88.4 | 2.2 | |
| ArrAttackModel=Yi-1.5-72B-Chat2026.04 | 90.9 | 86.8 | 7.8 | |
| ArrAttackModel=Mistral-7B-Instruct-v0.22026.04 | 87.1 | 82.9 | 8.1 | |
| HMNSModel=Qwen2.5-14B-Instruct2026.04 | 86.9 | 80.9 | 2.2 | |
| ArrAttackModel=Qwen2.5-14B-Instruct2026.04 | 81.2 | 75.4 | 8.7 | |
| AutoDANModel=Yi-1.5-72B-Chat2026.04 | 67.5 | 61.9 | 13 | |
| AutoDANModel=Mistral-7B-Instruct-v0.22026.04 | 64.9 | 59.4 | 13.5 | |
| AutoDANModel=Qwen2.5-14B-Instruct2026.04 | 58.2 | 52.8 | 14.3 | |
| FITDModel=Yi-1.5-72B-Chat2026.04 | 39.7 | 34.7 | 16.4 | |
| FITDModel=Mistral-7B-Instruct-v0.22026.04 | 36.8 | 31.7 | 17.2 | |
| FITDModel=Qwen2.5-14B-Instruct2026.04 | 34.3 | 29.6 | 17.9 |