List your company

Vision-language Math Reasoning on MathVista (Avg@32 (%))

71.8Avg@32MHPO
67.6468.7269.870.88Mar 14, 2026

Evaluation results

MethodLinks
MHPOBackbone=Qwen2.5-VL-7B-Instruct2026.0371.8
DAPOBackbone=Qwen2.5-VL-7B-Instruct2026.0371.2
SAPOBackbone=Qwen2.5-VL-7B-Instruct2026.0370.7
GSPOBackbone=Qwen2.5-VL-7B-Instruct2026.0369.1
GRPOBackbone=Qwen2.5-VL-7B-Instruct2026.0368.6
BaselineBackbone=Qwen2.5-VL-7B-Instruct2026.0367.8