VLM Jailbreak Detection on VLM Jailbreak Average (MM-SafetyBench, JailBreakV-28K, FigImg)
0.94Average AUROCMTK
Evaluation Results
| Method | Links | |
|---|---|---|
| MTKVLM=LLaVa2026.06 | 0.94 | |
| HiddenDetectVLM=LLaVa2026.06 | 0.928 | |
| MTKVLM=Qwen-VL2026.06 | 0.924 | |
| HiddenDetectVLM=Qwen-VL2026.06 | 0.878 | |
| ECSOVLM=LLaVa2026.06 | 0.825 | |
| JailGuardVLM=Qwen-VL2026.06 | 0.809 | |
| MirrorCheckVLM=Qwen-VL2026.06 | 0.708 | |
| ECSOVLM=Qwen-VL2026.06 | 0.683 | |
| MirrorCheckVLM=LLaVa2026.06 | 0.669 | |
| CIDERVLM=LLaVa2026.06 | 0.655 | |
| CIDERVLM=Qwen-VL2026.06 | 0.655 | |
| JailDAMVLM=LLaVa2026.06 | 0.627 | |
| JailDAMVLM=Qwen-VL2026.06 | 0.593 | |
| JailGuardVLM=LLaVa2026.06 | 0.547 |