Visual Mathematical Reasoning on FlowVerse Vision Centric RP+EI+OQ
32.6CoT ErrorQwen-VL-Plus
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen-VL-Plus2025.03 | 32.6 | 26 | |
| Gemini-Pro2025.03 | 36.1 | 27.3 | |
| InfiMM-Math-7B2025.03 | 39.6 | 30.3 | |
| InternVL2.5-8B2025.03 | 41 | 35.9 | |
| MathLLaVA-13B2025.03 | 41.7 | 35.9 | |
| Qwen-VL-Max2025.03 | 42.7 | 33.2 | |
| VLM-R1-7B†2025.03 | 46.1 | 37.8 | |
| MultiMath-7B2025.03 | 47.2 | 40.4 | |
| SVE-Math-Qwen2.5-7B2025.03 | 48.9 | 44.2 | |
| Qwen2.5-VL-7B2025.03 | 50.8 | 46.3 | |
| Qwen2-VL-72B2025.03 | 50.9 | 42.1 | |
| MathFlowBackbone=Qwen2.5-VL-7B2025.03 | 54.5 | 51.6 | |
| Claude-3.5-Sonnet2025.03 | 56.5 | 48 | |
| GPT-4o-mini2025.03 | 57.4 | 50.1 | |
| GPT-4o2025.03 | 58.6 | 52.6 | |
| GPT-4V2025.03 | 61.8 | 46.3 | |
| InternVL2.5-78B2025.03 | 63 | 58.8 | |
| Gemini-2.5-pro2025.03 | 65.7 | 57.9 | |
| GPT-52025.03 | 67.1 | 61.7 | |
| MathFlowBackbone=GPT-52025.03 | 70.2 | 66.7 |