Jailbreak Detection on Jailbreak V28K
96.4AUROCMTK
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MTKVLM=Qwen-VL2026.06 | 96.4 | — | — | — | — | — | |
| MTKVLM=LLaVa2026.06 | 94.9 | — | — | — | — | — | |
| HiddenDetectVLM=LLaVa2026.06 | 93.8 | — | — | — | — | — | |
| HiddenDetectVLM=Qwen-VL2026.06 | 88.1 | — | — | — | — | — | |
| ECSOVLM=LLaVa2026.06 | 86.6 | — | — | — | — | — | |
| JailGuardVLM=Qwen-VL2026.06 | 82.3 | — | — | — | — | — | |
| ECSOVLM=Qwen-VL2026.06 | 71.2 | — | — | — | — | — | |
| MirrorCheckVLM=Qwen-VL2026.06 | 62.8 | — | — | — | — | — | |
| JailDAMVLM=Qwen-VL2026.06 | 62.7 | — | — | — | — | — | |
| MirrorCheckVLM=LLaVa2026.06 | 61.1 | — | — | — | — | — | |
| CIDERVLM=LLaVa2026.06 | 60.7 | — | — | — | — | — | |
| CIDERVLM=Qwen-VL2026.06 | 60.7 | — | — | — | — | — | |
| JailDAMVLM=LLaVa2026.06 | 59.8 | — | — | — | — | — | |
| JailGuardVLM=LLaVa2026.06 | 58.6 | — | — | — | — | — | |
| Sentra-GuardPrompt Types=Code-based, role-play, narrative.2025.10 | — | 99.91 | 99.93 | 99.88 | 99.9 | 0.9 | |
| Sentra-GuardPrompt Types=Misinformation, cyberattacks, financial scams, and hate speech2025.10 | — | 99.98 | 100 | 99.97 | 99.98 | 0.4 |