Visual Question Answering on SGG-VQA Anatomy
47.32Avg@5GRPO-GDQA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GRPO-GDQAModel=Qwen-2.5-VL-7B2025.12 | 47.32 | 47.1 | 47.3 | |
| GRPO-GDQAModel=Qwen-2.5-VL-3B2025.12 | 43.36 | 42.8 | 43.5 | |
| GRPO-ASC-LearningModel=Qwen-2.5-VL-7B2025.12 | 42.18 | 43.8 | 44.2 | |
| SFT+GRPOModel=Qwen-2.5-VL-7B2025.12 | 41.28 | 42.3 | 42.9 | |
| GPGModel=Qwen-2.5-VL-7B2025.12 | 41.09 | 40.1 | 41.2 | |
| DAPOModel=Qwen-2.5-VL-3B2025.12 | 40.28 | 42.4 | 39.2 | |
| DAPOModel=Qwen-2.5-VL-7B2025.12 | 40.28 | 39.9 | 40.7 | |
| GRPO-ASC-LearningModel=Qwen-2.5-VL-3B2025.12 | 40.19 | 42.7 | 40.1 | |
| GPGModel=Qwen-2.5-VL-3B2025.12 | 40.17 | 41.4 | 40.4 | |
| GRPOModel=Qwen-2.5-VL-7B2025.12 | 38.82 | 39.7 | 38.5 | |
| GRPOModel=Qwen-2.5-VL-3B2025.12 | 36.5 | 36.1 | 37.2 | |
| SFTModel=Qwen-2.5-VL-7B2025.12 | 36.46 | 36.1 | 38.4 | |
| CoT SFT+GRPOModel=Qwen-2.5-VL-7B2025.12 | 34.78 | 33.2 | 34.2 | |
| SFTModel=Qwen-2.5-VL-3B2025.12 | 34.1 | 35.9 | 32.1 | |
| SFT+GRPOModel=Qwen-2.5-VL-3B2025.12 | 33.64 | 36.3 | 33.2 | |
| CoT SFT+GRPOModel=Qwen-2.5-VL-3B2025.12 | 31.32 | 34.6 | 34.2 | |
| CoT-SFTModel=Qwen-2.5-VL-7B2025.12 | 31.25 | 33.4 | 33.7 | |
| Zero-ShotModel=Qwen-2.5-VL-7B2025.12 | 29 | 29.6 | 29.1 | |
| Zero-ShotModel=Qwen-2.5-VL-3B2025.12 | 26.7 | 25.9 | 26.9 | |
| CoT-SFTModel=Qwen-2.5-VL-3B2025.12 | 26.32 | 24.7 | 25.8 |