Critical View of Safety on MedVidBench
91.4CVS AccuracyQwen2.5VL-7B_MedGRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5VL-7B_MedGRPOTraining=MedGRPO2025.12 | 91.4 | |
| Qwen2.5VL-7B_SFTTraining=SFT baseline2025.12 | 89.4 | |
| Qwen2.5VL-7BType=off-the-shelf2025.12 | 10.5 | |
| Gemini-2.5-flashType=off-the-shelf2025.12 | 10.1 | |
| GPT-4.1Type=off-the-shelf2025.12 | 1.8 |