List your company
ResearchFormerly WizWand

Mathematical Reasoning on DynaMath DMath

56.9AccuracyCPPO
32.77239.03645.351.564Jan 1, 2026
Updated 8mo ago

Evaluation results

MethodLinks
CPPOBackbone=Qwen2.5-VL-7B, Sampling Strategy=avg@8, Inference Temperature=12026.0156.9
NoisyRolloutBackbone=Qwen2.5-VL-7B, Sampling Strategy=avg@8, Inference Temperature=12026.0155.9
PerceptionR1Backbone=Qwen2.5-VL-7B, Sampling Strategy=avg@8, Inference Temperature=12026.0155.8
GRPOBackbone=Qwen2.5-VL-7B, Sampling Strategy=avg@8, Inference Temperature=12026.0155.6
PAPOBackbone=Qwen2.5-VL-7B, Sampling Strategy=avg@8, Inference Temperature=12026.0154.7
Vision-MattersBackbone=Qwen2.5-VL-7B, Sampling Strategy=avg@8, Inference Temperature=12026.0154.5
Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B, Sampling Strategy=avg@8, Inference Temperature=12026.0153.2
Vision-SR1Backbone=Qwen2.5-VL-7B, Sampling Strategy=avg@8, Inference Temperature=12026.0152.6
Look-BackBackbone=Qwen2.5-VL-7B, Sampling Strategy=avg@8, Inference Temperature=12026.0152.5
CPPOBackbone=Qwen2.5-VL-3B, Sampling Strategy=avg@8, Inference Temperature=12026.0148.9
GRPOBackbone=Qwen2.5-VL-3B, Sampling Strategy=avg@8, Inference Temperature=12026.0145.7
PAPOBackbone=Qwen2.5-VL-3B, Sampling Strategy=avg@8, Inference Temperature=12026.0145.4
OpenVLThinkerBackbone=Qwen2.5-VL-7B, Sampling Strategy=avg@8, Inference Temperature=12026.0143.9
Gemini-2.0-FlashSampling Strategy=avg@8, Inference Temperature=12026.0142.1
Visionary-R1Backbone=Qwen2.5-VL-3B, Sampling Strategy=avg@8, Inference Temperature=12026.0141.2
OpenVLThinkerBackbone=Qwen2.5-VL-3B, Sampling Strategy=avg@8, Inference Temperature=12026.0135.6
GPT4-oSampling Strategy=avg@8, Inference Temperature=12026.0134.5
Qwen2.5-VL-3BBackbone=Qwen2.5-VL-3B, Sampling Strategy=avg@8, Inference Temperature=12026.0133.7