Multimodal Scientific Reasoning on WeMath
71.77AccuracyCSRS
Evaluation Results
| Method | Links | |
|---|---|---|
| CSRSUnsupervised=true, Training Data=Geometry3K2026.04 | 71.77 | |
| CSRSUnsupervised=true, Training Data=MMR12026.04 | 70.53 | |
| CSRSUnsupervised=true, Training Data=GeoQA2026.04 | 69.32 | |
| MM-UPTUnsupervised=true, Training Data=MMR12026.04 | 69.14 | |
| MM-UPTUnsupervised=true, Training Data=Geometry3K2026.04 | 68.49 | |
| MM-UPTUnsupervised=true, Training Data=GeoQA2026.04 | 68.25 | |
| Qwen2.5-VL-7BUnsupervised=false2026.04 | 67.65 | |
| Qwen2.5-VL-3BUnsupervised=false2026.04 | 57.53 |