Jailbreak Detection on AdvBench Llama2-7B
97.9AUROCGradSafe
Evaluation Results
| Method | Links | |
|---|---|---|
| GradSafeAttack=AutoDAN2026.06 | 97.9 | |
| MTKAttack=SAA2026.06 | 96.1 | |
| MTKAttack=DrAttack2026.06 | 95.9 | |
| MTKAttack=TAP2026.06 | 95.6 | |
| MTKAttack=PAIR2026.06 | 95.3 | |
| SelfDefendAttack=SAA2026.06 | 94.6 | |
| SelfDefendAttack=AutoDAN2026.06 | 94.5 | |
| MTKAttack=PAP2026.06 | 94.4 | |
| GradSafeAttack=TAP2026.06 | 94.1 | |
| MTKAttack=Avg.2026.06 | 94 | |
| SelfDefendAttack=GCG2026.06 | 93.9 | |
| GradSafeAttack=PAIR2026.06 | 93.6 | |
| HSFAttack=DrAttack2026.06 | 92.9 | |
| MTKAttack=JailJudge2026.06 | 92.9 | |
| MTKAttack=GCG2026.06 | 92.9 | |
| MTKAttack=AutoDAN2026.06 | 92.7 | |
| HiddenDetectAttack=AutoDAN2026.06 | 92.7 | |
| MTKAttack=IJP2026.06 | 92.4 | |
| HiddenDetectAttack=JailJudge2026.06 | 92.4 | |
| MTKAttack=Zulu2026.06 | 92.2 | |
| HiddenDetectAttack=Zulu2026.06 | 92.1 | |
| GradSafeAttack=GCG2026.06 | 90.4 | |
| SelfDefendAttack=PAIR2026.06 | 89.3 | |
| SelfDefendAttack=IJP2026.06 | 89.1 | |
| SaPAttack=Zulu2026.06 | 89 | |
| GradSafeAttack=SAA2026.06 | 88.6 | |
| HSFAttack=IJP2026.06 | 87.6 | |
| SelfDefendAttack=Zulu2026.06 | 87.6 | |
| GradSafeAttack=IJP2026.06 | 86.3 | |
| SelfDefendAttack=TAP2026.06 | 86.1 | |
| HiddenDetectAttack=TAP2026.06 | 85.8 | |
| GradSafeAttack=JailJudge2026.06 | 85.7 | |
| HiddenDetectAttack=PAIR2026.06 | 85.6 | |
| HSFAttack=TAP2026.06 | 85.3 | |
| SelfDefendAttack=Avg.2026.06 | 84.8 | |
| SaPAttack=PAP2026.06 | 84.4 | |
| SmoothLLMAttack=JailJudge2026.06 | 83.5 | |
| HSFAttack=SAA2026.06 | 82.5 | |
| GradSafeAttack=Avg.2026.06 | 81.3 | |
| SelfDefendAttack=PAP2026.06 | 80.1 | |
| HSFAttack=PAIR2026.06 | 79 | |
| HiddenDetectAttack=GCG2026.06 | 78.6 | |
| SaPAttack=AutoDAN2026.06 | 78 | |
| GradSafeAttack=DrAttack2026.06 | 77.6 | |
| HiddenDetectAttack=Avg.2026.06 | 77.6 | |
| GradCuffAttack=TAP2026.06 | 77.2 | |
| HiddenDetectAttack=DrAttack2026.06 | 76.3 | |
| GradCuffAttack=AutoDAN2026.06 | 75.8 | |
| SmoothLLMAttack=Zulu2026.06 | 74.5 | |
| SmoothLLMAttack=AutoDAN2026.06 | 73.3 | |
| GradSafeAttack=Zulu2026.06 | 73.3 | |
| HSFAttack=Avg.2026.06 | 72 | |
| GradCuffAttack=PAIR2026.06 | 71.9 | |
| SmoothLLMAttack=SAA2026.06 | 71.8 | |
| SelfDefendAttack=DrAttack2026.06 | 71.3 | |
| GradCuffAttack=IJP2026.06 | 71.2 | |
| HSFAttack=GCG2026.06 | 70.2 | |
| HSFAttack=PAP2026.06 | 69.5 | |
| SaPAttack=IJP2026.06 | 68.5 | |
| HiddenDetectAttack=IJP2026.06 | 67.3 | |
| GradCuffAttack=JailJudge2026.06 | 67.3 | |
| SaPAttack=GCG2026.06 | 65.3 | |
| GradCuffAttack=Avg.2026.06 | 64.7 | |
| GradCuffAttack=SAA2026.06 | 63.7 | |
| SmoothLLMAttack=PAP2026.06 | 63.4 | |
| SaPAttack=Avg.2026.06 | 63.3 | |
| SelfDefendAttack=JailJudge2026.06 | 61.9 | |
| HiddenDetectAttack=SAA2026.06 | 61.5 | |
| GradCuffAttack=DrAttack2026.06 | 57.6 | |
| GradCuffAttack=GCG2026.06 | 57 | |
| HSFAttack=JailJudge2026.06 | 56.1 | |
| SmoothLLMAttack=Avg.2026.06 | 55.8 | |
| SmoothLLMAttack=GCG2026.06 | 55.1 | |
| GradCuffAttack=Zulu2026.06 | 54.5 | |
| HSFAttack=AutoDAN2026.06 | 51.8 | |
| SmoothLLMAttack=TAP2026.06 | 51.6 | |
| GradCuffAttack=PAP2026.06 | 51.2 | |
| SaPAttack=PAIR2026.06 | 50.9 | |
| SaPAttack=TAP2026.06 | 50.2 | |
| SaPAttack=DrAttack2026.06 | 50 | |
| SaPAttack=JailJudge2026.06 | 49.7 | |
| SaPAttack=SAA2026.06 | 46.6 | |
| HSFAttack=Zulu2026.06 | 44.8 | |
| SmoothLLMAttack=PAIR2026.06 | 44.4 | |
| HiddenDetectAttack=PAP2026.06 | 43.2 | |
| GradSafeAttack=PAP2026.06 | 25.2 | |
| SmoothLLMAttack=IJP2026.06 | 22.4 | |
| SmoothLLMAttack=DrAttack2026.06 | 18 |