Jailbreak Attack Success Evaluation on HADES SD
2.8ASRJRS-Rem
Evaluation Results
| Method | Links | |
|---|---|---|
| JRS-RemVLM=ShareGPT4V-7B, Defense=JRS-Rem2026.03 | 2.8 | |
| JRS-RemVLM=InternVL-Chat-19B, Defense=JRS-Rem2026.03 | 5.6 | |
| CMRMVLM=ShareGPT4V-7B, Defense=CMRM2026.03 | 5.7 | |
| JRS-RemVLM=LLaVA-1.5-7B, Defense=JRS-Rem2026.03 | 12.2 | |
| AdaShieldVLM=ShareGPT4V-7B, Defense=AdaShield2026.03 | 12.5 | |
| AdaShieldVLM=InternVL-Chat-19B, Defense=AdaShield2026.03 | 13.7 | |
| ECSOVLM=InternVL-Chat-19B, Defense=ECSO2026.03 | 19.4 | |
| ShiftDCVLM=InternVL-Chat-19B, Defense=ShiftDC2026.03 | 24.1 | |
| ShiftDCVLM=ShareGPT4V-7B, Defense=ShiftDC2026.03 | 26.4 | |
| AdaShieldVLM=LLaVA-1.5-7B, Defense=AdaShield2026.03 | 28.4 | |
| ECSOVLM=LLaVA-1.5-7B, Defense=ECSO2026.03 | 30 | |
| CMRMVLM=InternVL-Chat-19B, Defense=CMRM2026.03 | 35.6 | |
| ECSOVLM=ShareGPT4V-7B, Defense=ECSO2026.03 | 36.1 | |
| InternVL-Chat-19B (Baseline)VLM=InternVL-Chat-19B, Defense=None2026.03 | 38.9 | |
| ShareGPT4V-7B (Baseline)VLM=ShareGPT4V-7B, Defense=None2026.03 | 58.1 | |
| ShiftDCVLM=LLaVA-1.5-7B, Defense=ShiftDC2026.03 | 67.2 | |
| CMRMVLM=LLaVA-1.5-7B, Defense=CMRM2026.03 | 73.8 | |
| LLaVA-1.5-7B (Baseline)VLM=LLaVA-1.5-7B, Defense=None2026.03 | 77.3 |