Jailbreak Detection on AdvBench Mistral-7B
99AUROCMTK
Evaluation Results
| Method | Links | |
|---|---|---|
| MTKAttack=AutoDAN2026.06 | 99 | |
| MTKAttack=Avg.2026.06 | 95.3 | |
| SelfDefendAttack=AutoDAN2026.06 | 94.5 | |
| HiddenDetectAttack=AutoDAN2026.06 | 85.8 | |
| SelfDefendAttack=Avg.2026.06 | 84.8 | |
| GradCuffAttack=AutoDAN2026.06 | 82 | |
| HSFAttack=Avg.2026.06 | 80.8 | |
| HiddenDetectAttack=Avg.2026.06 | 78.9 | |
| HSFAttack=AutoDAN2026.06 | 71.7 | |
| SaPAttack=Avg.2026.06 | 63.5 | |
| GradCuffAttack=Avg.2026.06 | 60 | |
| SmoothLLMAttack=Avg.2026.06 | 53.7 | |
| SaPAttack=AutoDAN2026.06 | 50.5 | |
| GradSafeAttack=Avg.2026.06 | 47.3 | |
| GradSafeAttack=AutoDAN2026.06 | 43.1 | |
| SmoothLLMAttack=AutoDAN2026.06 | 35.7 |