Safety Evaluation on HarmBench FigStep
1Safety Score (1-ASR)RESTA
Evaluation Results
| Method | Links | |
|---|---|---|
| RESTABackbone Model=Lingshu-7B, Alignment Method=RESTA2025.12 | 1 | |
| PSIBackbone Model=HuatuoGPT-Vision-7B, Alignment Method=Parameter-Space Intervention2025.12 | 1 | |
| RESTABackbone Model=HuatuoGPT-Vision-7B, Alignment Method=RESTA2025.12 | 0.99 | |
| PSIBackbone Model=Lingshu-7B, Alignment Method=Parameter-Space Intervention2025.12 | 0.9675 | |
| ModelMergeBackbone Model=HuatuoGPT-Vision-7B, Alignment Method=ModelMerge2025.12 | 0.3135 | |
| ModelMergeBackbone Model=Lingshu-7B, Alignment Method=ModelMerge2025.12 | 0.2575 | |
| OriginalBackbone Model=Lingshu-7B, Alignment Method=None2025.12 | 0.2375 | |
| OriginalBackbone Model=HuatuoGPT-Vision-7B, Alignment Method=None2025.12 | 0.2025 |