Dynamic Mathematical Reasoning on DynaMath (test)
64.8AccuracyClaude-3.5 Sonnet
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Claude-3.5 SonnetModel Category=Closed-Source Models2026.01 | 64.8 | — | |
| GPT-4oModel Category=Closed-Source Models2026.01 | 63.7 | — | |
| GPRO-7BBase Model Scale=7B, Base Architecture=Qwen2.5-VL2026.01 | 59.2 | 158.7 | |
| FAST-7BBase Model Scale=7B, Base Architecture=Qwen2.5-VL2026.01 | 58.3 | 164.8 | |
| Qwen2.5-VL-7BBase Model Scale=7B, Base Architecture=Qwen2.5-VL2026.01 | 58 | 273.3 | |
| MM-R1Base Model Scale=7B, Base Architecture=Qwen2.5-VL2026.01 | 57.5 | 254.2 | |
| OpenVLThinkerBase Model Scale=7B, Base Architecture=Qwen2.5-VL2026.01 | 57.4 | 382.1 | |
| GPRO-3BBase Model Scale=3B, Base Architecture=Qwen2.5-VL2026.01 | 55.1 | 195.3 | |
| FAST-3BBase Model Scale=3B, Base Architecture=Qwen2.5-VL2026.01 | 54.4 | 213.7 | |
| R1-OneVisionBase Model Scale=7B, Base Architecture=Qwen2.5-VL2026.01 | 53.5 | 560.6 | |
| LMM-R1Base Model Scale=3B, Base Architecture=Qwen2.5-VL2026.01 | 53.1 | 341.6 | |
| Qwen2.5-VL-3BBase Model Scale=3B, Base Architecture=Qwen2.5-VL2026.01 | 48.2 | 270.9 | |
| MulberryBase Model Scale=7B, Base Architecture=Qwen2-VL2026.01 | 46.8 | 273.3 | |
| Curr-ReFTBase Model Scale=3B, Base Architecture=Qwen2.5-VL2026.01 | 43.8 | 146.4 | |
| Qwen2-VL-7BBase Model Scale=7B, Base Architecture=Qwen2-VL2026.01 | 39.8 | 298.4 |