List your company
ResearchFormerly WizWand

Multimodal Mathematical Reasoning on DynaMath

66.4DynaMath ScoreVision-R1-72B*
33.32841.91450.559.086Mar 9, 2025May 18, 2025Jul 28, 2025Oct 7, 2025Dec 17, 2025Feb 26, 2026May 8, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Vision-R1-72B*Params.=72B, additional data during RL training=true2025.0366.4
Vision-R1-32B*Params.=32B, additional data during RL training=true2025.0365.6
GPT-4oParams.=-2025.0364.9
Claude-3.5 SonnetParams.=-2025.0362.5
Qwen2.5-VL-72BParams.=72B2025.0361.2
PRCO-7BBackbone=Qwen2.5-VL-7B2026.0559.36
Qwen2.5-VL-7B + SRPOBackbone=Qwen2.5-VL-7B2026.0558.22
VPPO-7BBackbone=Qwen2.5-VL-7B2026.0556.88
PAPO-D-7BBackbone=Qwen2.5-VL-7B2026.0556.54
Vision-R1-7BParams.=7B2025.0356.3
Qwen2.5-VL-32BParams.=32B2025.0355.5
Qwen2.5-VL-7B + DAPOBackbone=Qwen2.5-VL-7B2026.0555.46
Perception-R1-7BBackbone=Qwen2.5-VL-7B2026.0555.36
VL-Rethinker-7BBackbone=Qwen2.5-VL-7B2026.0555.14
Vision-Matters-7BBackbone=Qwen2.5-VL-7B2026.0555.09
Qwen2.5-VL-7B + GRPOBackbone=Qwen2.5-VL-7B2026.0555.09
PAPO-G-7BBackbone=Qwen2.5-VL-7B2026.0553.11
Vision-SR1-7BBackbone=Qwen2.5-VL-7B2026.0552.71
ThinkLite-VL-7BBackbone=Qwen2.5-VL-7B2026.0551.79
Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2026.0551.29
Qwen2.5-VL-7BParams.=7B2025.0350.7
Qwen2.5-VL-3B + SRPOBackbone=Qwen2.5-VL-3B2026.0548.54
PRCO-3BBackbone=Qwen2.5-VL-3B2026.0546.31
Vision-SR1-3BBackbone=Qwen2.5-VL-3B2026.0545.34
PAPO-D-3BBackbone=Qwen2.5-VL-3B2026.0545.16
Qwen2.5-VL-3B + GRPOBackbone=Qwen2.5-VL-3B2026.0545.06
PAPO-G-3BBackbone=Qwen2.5-VL-3B2026.0544.89
Qwen2.5-VL-3B + DAPOBackbone=Qwen2.5-VL-3B2026.0544.53
Qwen2.5-VL-3BBackbone=Qwen2.5-VL-3B2026.0543.15
InternVL2.5-78BParams.=78B2025.0341.4
LLaVA-CoT-11BParams.=11B2025.0334.6