Visual Question Answering on OmniMedVQA MI
83.6Avg@5GRPO-GDQA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GRPO-GDQAModel=Qwen-2.5-VL-7B2025.12 | 83.6 | 83.72 | 83.52 | |
| GPGModel=Qwen-2.5-VL-7B2025.12 | 80.89 | 80.65 | 81.07 | |
| DAPOModel=Qwen-2.5-VL-7B2025.12 | 80.16 | 81.1 | 82 | |
| GRPO-GDQAModel=Qwen-2.5-VL-3B2025.12 | 79.81 | 82.38 | 82.77 | |
| SFTModel=Qwen-2.5-VL-7B2025.12 | 79.16 | 80.27 | 80.08 | |
| SFT+GRPOModel=Qwen-2.5-VL-7B2025.12 | 77.91 | 79.83 | 80.06 | |
| DAPOModel=Qwen-2.5-VL-3B2025.12 | 77.84 | 78.89 | 80.12 | |
| SFTModel=Qwen-2.5-VL-3B2025.12 | 77.39 | 77.39 | 77.97 | |
| GPGModel=Qwen-2.5-VL-3B2025.12 | 76.91 | 78.86 | 78.43 | |
| SFT+GRPOModel=Qwen-2.5-VL-3B2025.12 | 76.32 | 78.91 | 79.32 | |
| CoT SFT+GRPOModel=Qwen-2.5-VL-7B2025.12 | 76.29 | 76.01 | 77.49 | |
| GRPOModel=Qwen-2.5-VL-7B2025.12 | 76.14 | 77.8 | 80.72 | |
| GRPOModel=Qwen-2.5-VL-3B2025.12 | 75.13 | 75.16 | 76.25 | |
| CoT-SFTModel=Qwen-2.5-VL-7B2025.12 | 74.13 | 77.78 | 78.31 | |
| CoT-SFTModel=Qwen-2.5-VL-3B2025.12 | 68.93 | 70.74 | 71.02 | |
| CoT SFT+GRPOModel=Qwen-2.5-VL-3B2025.12 | 67.43 | 68.31 | 68.45 | |
| Zero-ShotModel=Qwen-2.5-VL-7B2025.12 | 60.77 | 62.26 | 69.16 | |
| Zero-ShotModel=Qwen-2.5-VL-3B2025.12 | 60.08 | 59.39 | 68.01 |