Visual Math on MathVista
80.9Pass@5ViGOS
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| ViGOSBackbone=Qwen2.5-VL 7B2026.06 | 80.9 | — | — | — | — | — | 58.78 | |
| OPSDBackbone=Qwen2.5-VL 7B2026.06 | 80.5 | — | — | — | — | — | 61.54 | |
| BaselineBackbone=Qwen2.5-VL 7B2026.06 | 79.2 | — | — | — | — | — | 60.9 | |
| Vision-R1-7BMethod Category=RL Method2026.06 | 77.3 | — | — | — | — | — | 63.92 | |
| ViGOSBackbone=Qwen2.5-VL 3B2026.06 | 74 | — | — | — | — | — | 43.5 | |
| OPSDBackbone=Qwen2.5-VL 3B2026.06 | 72.8 | — | — | — | — | — | 43.84 | |
| Visionary-R1-3BMethod Category=RL Method2026.06 | 72 | — | — | — | — | — | 57.18 | |
| BaselineBackbone=Qwen2.5-VL 3B2026.06 | 65.4 | — | — | — | — | — | 35.08 | |
| Base Modelevaluation_mode=Out-of-Distribution (OOD), training_protocol=Transfer from ChartQA2026.02 | — | 45 | 78 | — | — | — | — | |
| Claude4-Sonnet2026.02 | — | — | — | — | 72.4 | 64.1 | — | |
| DeepEyes2026.02 | — | — | — | — | 70.1 | — | — | |
| DeepEyesV22026.02 | — | — | — | — | 38.1 | — | — | |
| DPEBackbone=Qwen3-VL-8B-Instruct, Iteration=32026.02 | — | — | — | — | 76.2 | 64.39 | — | |
| GPT-4o2026.02 | — | — | — | — | 63.8 | 56.1 | — | |
| GPT5-Mini2026.02 | — | — | — | — | 59.6 | 53.8 | — | |
| GRPOevaluation_mode=Out-of-Distribution (OOD), training_protocol=Transfer from ChartQA2026.02 | — | 45 | 78 | — | — | — | — | |
| MIGevaluation_mode=Out-of-Distribution (OOD), training_protocol=Transfer from ChartQA2026.02 | — | 45 | 71 | 0 | — | — | — | |
| Qwen2.5-VL-72BParameters=72B2026.02 | — | — | — | — | 74.8 | 61.9 | — |