Jailbreak Detection on JailDAM Overall evaluation setup
0.9936AUROCMahal-OOD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Mahal-OODModel=LLaVA, Uses target model hidden states=true2025.12 | 0.9936 | 0.9969 | |
| Mahal-OOD*Model=FLAVA, Uses target model hidden states=false2025.12 | 0.977 | 0.9889 | |
| KNN-OODModel=LLaVA, Uses target model hidden states=true2025.12 | 0.9638 | 0.9819 | |
| JailDAM*Model=CLIP, Uses target model hidden states=false2025.12 | 0.955 | 0.953 | |
| KNN-OOD*Model=FLAVA, Uses target model hidden states=false2025.12 | 0.8867 | 0.95 | |
| GradSafeModel=LLaVA2025.12 | 0.8513 | 0.8166 | |
| HiddenDetectModel=LLaVA2025.12 | 0.805 | 0.8056 | |
| LLaVaGuardModel=Qwen2025.12 | 0.7551 | 0.8412 | |
| VLGuardModel=LLaVA2025.12 | 0.6096 | 0.6782 |