Jailbreak Defense on HADES
3.6ASRJRS-Rem
Evaluation results
| Method | Links | |
|---|---|---|
| JRS-RemModel=InternVL-Chat-19B2026.03 | 3.6 | |
| JRS-RemModel=ShareGPT4V-7B2026.03 | 9.2 | |
| CMRMModel=ShareGPT4V-7B2026.03 | 10.8 | |
| AdaShieldModel=ShareGPT4V-7B2026.03 | 11.6 | |
| AdaShieldModel=InternVL-Chat-19B2026.03 | 20.5 | |
| CMRMModel=InternVL-Chat-19B2026.03 | 25.6 | |
| ECSOModel=InternVL-Chat-19B2026.03 | 30.1 | |
| AdaShieldBackbone=LLaVA-1.5-7B2026.03 | 35.2 | |
| AdaShieldModel=LLaVA-1.5-7B2026.03 | 35.2 | |
| JRS-RemBackbone=LLaVA-1.5-7B2026.03 | 35.3 | |
| JRS-RemModel=LLaVA-1.5-7B2026.03 | 35.3 | |
| ECSOBackbone=LLaVA-1.5-7B2026.03 | 44.1 | |
| ECSOModel=LLaVA-1.5-7B2026.03 | 44.1 | |
| ShiftDCModel=ShareGPT4V-7B2026.03 | 47.3 | |
| ECSOModel=ShareGPT4V-7B2026.03 | 49 | |
| CMRMBackbone=LLaVA-1.5-7B2026.03 | 52.6 | |
| CMRMModel=LLaVA-1.5-7B2026.03 | 52.6 | |
| ShiftDCModel=InternVL-Chat-19B2026.03 | 57.4 | |
| ShiftDCBackbone=LLaVA-1.5-7B2026.03 | 60.4 | |
| ShiftDCModel=LLaVA-1.5-7B2026.03 | 60.4 | |
| ShareGPT4V-7BDefense=None2026.03 | 66.8 | |
| LLaVA-1.5-7BDefense=None2026.03 | 67 | |
| LLaVA-1.5-7BDefense=None2026.03 | 67 | |
| InternVL-Chat-19BDefense=None2026.03 | 67.7 |