List your company

Jailbreak Detection on JailDAM Overall evaluation setup

0.9936AUROCMahal-OOD
0.594240.697920.80160.90528Dec 12, 2025

Evaluation results

MethodLinks
Mahal-OODModel=LLaVA, Uses target model hidden states=true2025.120.99360.9969
Mahal-OOD*Model=FLAVA, Uses target model hidden states=false2025.120.9770.9889
KNN-OODModel=LLaVA, Uses target model hidden states=true2025.120.96380.9819
JailDAM*Model=CLIP, Uses target model hidden states=false2025.120.9550.953
KNN-OOD*Model=FLAVA, Uses target model hidden states=false2025.120.88670.95
GradSafeModel=LLaVA2025.120.85130.8166
HiddenDetectModel=LLaVA2025.120.8050.8056
LLaVaGuardModel=Qwen2025.120.75510.8412
VLGuardModel=LLaVA2025.120.60960.6782