Geometry Problem Solving on Geometry
50.6Average ScoreDAPO
Evaluation Results
| Method | Links | |
|---|---|---|
| DAPOBackbone=Qwen-VL-3B, Finetuning=true, Runtime=11.2h, Rollouts=1438k2026.05 | 50.6 | |
| POPOBackbone=Qwen-VL-3B, Finetuning=true, Runtime=6.8h, Rollouts=492k2026.05 | 50 | |
| MoPPSBackbone=Qwen-VL-3B, Finetuning=true, Runtime=7.4h, Rollouts=492k2026.05 | 48.9 | |
| ARPOBackbone=Qwen-VL-3B, Finetuning=true, Runtime=6.5h, Rollouts=492k2026.05 | 45.7 | |
| GRPOBackbone=Qwen-VL-3B, Finetuning=true, Runtime=6.8h, Rollouts=492k2026.05 | 44.9 | |
| Qwen-VL-3B2026.05 | 24.6 |