Jailbreak Attack Success Evaluation on MM-SafetyBench SD
80.6ASRLLaVA-1.5-7B (Baseline)
Evaluation results
| Method | Links | |
|---|---|---|
| LLaVA-1.5-7B (Baseline)VLM=LLaVA-1.5-7B2026.03 | 80.6 | |
| InternVL-Chat-19B (Baseline)VLM=InternVL-Chat-19B2026.03 | 71.7 | |
| ShareGPT4V-7B (Baseline)VLM=ShareGPT4V-7B2026.03 | 64.3 | |
| CMRMVLM=LLaVA-1.5-7B2026.03 | 49.2 | |
| ShiftDCVLM=LLaVA-1.5-7B2026.03 | 44.3 | |
| ShiftDCVLM=InternVL-Chat-19B2026.03 | 41.4 | |
| AdaShieldVLM=LLaVA-1.5-7B2026.03 | 32.6 | |
| AdaShieldVLM=InternVL-Chat-19B2026.03 | 28.8 | |
| ECSOVLM=InternVL-Chat-19B2026.03 | 25.1 | |
| CMRMVLM=InternVL-Chat-19B2026.03 | 19.7 | |
| ECSOVLM=LLaVA-1.5-7B2026.03 | 17 | |
| ShiftDCVLM=ShareGPT4V-7B2026.03 | 16.1 | |
| ECSOVLM=ShareGPT4V-7B2026.03 | 13.7 | |
| JRS-RemVLM=LLaVA-1.5-7B2026.03 | 12.6 | |
| CMRMVLM=ShareGPT4V-7B2026.03 | 12.4 | |
| AdaShieldVLM=ShareGPT4V-7B2026.03 | 11.8 | |
| JRS-RemVLM=ShareGPT4V-7B2026.03 | 5.4 | |
| JRS-RemVLM=InternVL-Chat-19B2026.03 | 4.4 |