List your company

Jailbreak Detection on AdvBench Vicuna-7B

95.7AUROCMTK
19.67639.41359.1578.887Jun 5, 2026

Evaluation results

MethodLinks
MTKAttack=AutoDAN2026.0695.7
SelfDefendAttack=AutoDAN2026.0694.3
MTKAttack=Avg.2026.0692.3
HiddenDetectAttack=Avg.2026.0689.5
HiddenDetectAttack=AutoDAN2026.0687.5
SelfDefendAttack=Avg.2026.0685.5
HSFAttack=Avg.2026.0671
GradCuffAttack=Avg.2026.0666.8
GradCuffAttack=AutoDAN2026.0663.5
SaPAttack=AutoDAN2026.0658.4
SmoothLLMAttack=Avg.2026.0657.8
SmoothLLMAttack=AutoDAN2026.0653.5
SaPAttack=Avg.2026.0653.2
GradSafeAttack=AutoDAN2026.0649.7
GradSafeAttack=Avg.2026.0638
HSFAttack=AutoDAN2026.0622.6