Multimodal Math Reasoning on MathVerse (avg@8 accuracy)
68.58Avg@8 AccuracyPAPO_D
Evaluation Results
| Method | Links | |
|---|---|---|
| PAPO_DBackbone=Qwen2.5-VL, Model Size=7B2025.07 | 68.58 | |
| PAPO_GBackbone=Qwen2.5-VL, Model Size=7B2025.07 | 68.43 | |
| GRPOBackbone=Qwen2.5-VL, Model Size=7B2025.07 | 66.51 | |
| PAPO_DBackbone=Qwen2.5-VL, Model Size=3B2025.07 | 60.51 | |
| PAPO_GBackbone=Qwen2.5-VL, Model Size=3B2025.07 | 57.14 | |
| DAPOBackbone=Qwen2.5-VL, Model Size=3B2025.07 | 56.25 | |
| DAPOBackbone=Qwen2.5-VL, Model Size=7B2025.07 | 55.64 | |
| GRPOBackbone=Qwen2.5-VL, Model Size=3B2025.07 | 55.25 | |
| PAPO_GBackbone=Qwen3-VL (thinking), Model Size=2B2025.07 | 51.89 | |
| GRPOBackbone=Qwen3-VL (thinking), Model Size=2B2025.07 | 47.98 |