Vision-Dependent Multimodal Math Reasoning on MathVerseV Vision-centric
64.97Avg@8 AccuracyPAPO_G
Evaluation Results
| Method | Links | |
|---|---|---|
| PAPO_GBackbone=Qwen2.5-VL, Model Size=7B2025.07 | 64.97 | |
| PAPO_DBackbone=Qwen2.5-VL, Model Size=7B2025.07 | 64.87 | |
| GRPOBackbone=Qwen2.5-VL, Model Size=7B2025.07 | 61.71 | |
| PAPO_DBackbone=Qwen2.5-VL, Model Size=3B2025.07 | 57.72 | |
| PAPO_GBackbone=Qwen2.5-VL, Model Size=3B2025.07 | 53.95 | |
| DAPOBackbone=Qwen2.5-VL, Model Size=3B2025.07 | 53.09 | |
| GRPOBackbone=Qwen2.5-VL, Model Size=3B2025.07 | 52.26 | |
| DAPOBackbone=Qwen2.5-VL, Model Size=7B2025.07 | 51.04 | |
| PAPO_GBackbone=Qwen3-VL (thinking), Model Size=2B2025.07 | 50.05 | |
| GRPOBackbone=Qwen3-VL (thinking), Model Size=2B2025.07 | 45.41 |