Visual-Centric Reasoning on MME
86.3Accuracy@1DAPO
Evaluation results
| Method | Links | |
|---|---|---|
| DAPOModel Scale=Qwen2.5-VL-7B2026.06 | 86.3 | |
| DAPO + DyCo-RLModel Scale=Qwen2.5-VL-7B2026.06 | 86 | |
| SAPO + DyCo-RLModel Scale=Qwen2.5-VL-7B2026.06 | 86 | |
| SAPOModel Scale=Qwen2.5-VL-7B2026.06 | 85.6 | |
| GRPO + DyCo-RLModel Scale=Qwen2.5-VL-7B2026.06 | 84.9 | |
| GSPO + DyCo-RLModel Scale=Qwen2.5-VL-7B2026.06 | 84.8 | |
| GRPOModel Scale=Qwen2.5-VL-7B2026.06 | 83.3 | |
| Zero-ShotModel Scale=Qwen2.5-VL-7B2026.06 | 82.4 | |
| DAPO + DyCo-RLModel Scale=Qwen2.5-VL-3B2026.06 | 82.1 | |
| SAPOModel Scale=Qwen2.5-VL-3B2026.06 | 82.1 | |
| GSPO + DyCo-RLModel Scale=Qwen2.5-VL-3B2026.06 | 81.9 | |
| GSPOModel Scale=Qwen2.5-VL-7B2026.06 | 81.6 | |
| SAPO + DyCo-RLModel Scale=Qwen2.5-VL-3B2026.06 | 81.1 | |
| DAPOModel Scale=Qwen2.5-VL-3B2026.06 | 81 | |
| GRPO + DyCo-RLModel Scale=Qwen2.5-VL-3B2026.06 | 80.7 | |
| GRPOModel Scale=Qwen2.5-VL-3B2026.06 | 79.9 | |
| GSPOModel Scale=Qwen2.5-VL-3B2026.06 | 79.6 | |
| Zero-ShotModel Scale=Qwen2.5-VL-3B2026.06 | 76.9 |