List your company

Jailbreak Detection on FigStep (AUROC/AUPRC)

0.9955AUROCMahal-OOD
0.5605720.6734860.78640.899314Dec 12, 2025

Evaluation results

MethodLinks
Mahal-OODModel=LLaVA, Uses target model hidden states=true2025.120.99550.979
Mahal-OOD*Model=FLAVA, Uses target model hidden states=false2025.120.9820.9253
KNN-OODModel=LLaVA, Uses target model hidden states=true2025.120.98090.916
JailDAM*Model=CLIP, Uses target model hidden states=false2025.120.96080.9616
KNN-OOD*Model=FLAVA, Uses target model hidden states=false2025.120.90810.668
LLaVaGuardModel=Qwen2025.120.8360.7231
GradSafeModel=LLaVA2025.120.68040.237
VLGuardModel=LLaVA2025.120.61060.3817
HiddenDetectModel=LLaVA2025.120.57730.3238