Jailbreak Detection on FigStep (AUROC/AUPRC)
0.9955AUROCMahal-OOD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Mahal-OODModel=LLaVA, Uses target model hidden states=true2025.12 | 0.9955 | 0.979 | |
| Mahal-OOD*Model=FLAVA, Uses target model hidden states=false2025.12 | 0.982 | 0.9253 | |
| KNN-OODModel=LLaVA, Uses target model hidden states=true2025.12 | 0.9809 | 0.916 | |
| JailDAM*Model=CLIP, Uses target model hidden states=false2025.12 | 0.9608 | 0.9616 | |
| KNN-OOD*Model=FLAVA, Uses target model hidden states=false2025.12 | 0.9081 | 0.668 | |
| LLaVaGuardModel=Qwen2025.12 | 0.836 | 0.7231 | |
| GradSafeModel=LLaVA2025.12 | 0.6804 | 0.237 | |
| VLGuardModel=LLaVA2025.12 | 0.6106 | 0.3817 | |
| HiddenDetectModel=LLaVA2025.12 | 0.5773 | 0.3238 |