Multimodal Mathematical Reasoning on MathVista (GEO, ALG, GPS, MWP, ALL)
78.2Overall Accuracy (ALL)Vision-R1-72B*
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Vision-R1-72B*Params.=72B, additional data during RL training=true2025.03 | 78.2 | 88.3 | 86.8 | 89.4 | 79.6 | |
| Vision-R1-32B*Params.=32B, additional data during RL training=true2025.03 | 76.4 | 85.8 | 82.6 | 88 | 78.5 | |
| OpenAI O1Params.=-2025.03 | 73.9 | — | — | — | — | |
| Qwen2.5-VL-72BParams.=72B2025.03 | 73.5 | 77.8 | 77.9 | 78.8 | 74.7 | |
| Vision-R1-7BParams.=7B2025.03 | 73.5 | 80.3 | 79 | 83.2 | 80.6 | |
| Qwen2.5-VL-32BParams.=32B2025.03 | 72.9 | 72.8 | 73.7 | 75 | 72.6 | |
| InternVL2.5-78BParams.=78B2025.03 | 71.9 | 76.6 | 76.5 | 77.9 | 75.8 | |
| Qwen2.5-VL-7BParams.=7B2025.03 | 68.1 | 66.9 | 68.7 | 66.8 | 76.9 | |
| Claude-3.5 SonnetParams.=-2025.03 | 67.7 | — | — | — | — | |
| GPT-4oParams.=-2025.03 | 63.8 | — | — | — | — | |
| Mulberry-7BParams.=7B2025.03 | 63.1 | — | — | — | — | |
| URSA-8BParams.=8B2025.03 | 59.8 | — | — | 79.3 | 75.3 | |
| LLaVA-CoT-11BParams.=11B2025.03 | 54.8 | — | — | — | — | |
| Multimath-7BParams.=7B2025.03 | 50 | — | — | 66.8 | 61.8 | |
| GPT-4VParams.=-2025.03 | 49.9 | — | — | — | — | |
| Math-PUMA-Qwen2-7BParams.=7B2025.03 | 47.9 | 47.3 | — | 48.1 | — | |
| Math-LLaVA-13BParams.=13B2025.03 | 46.6 | 56.5 | 40.2 | 57.7 | 56.5 |