List your company

Vision-language math reasoning on MathVerse

56.7Avg@32 AccuracyMHPO
49.21251.15653.155.044Mar 14, 2026

Evaluation results

MethodLinks
MHPOBackbone=Qwen2.5-VL-7B-Instruct2026.0356.7
DAPOBackbone=Qwen2.5-VL-7B-Instruct2026.0356.5
SAPOBackbone=Qwen2.5-VL-7B-Instruct2026.0356.3
GRPOBackbone=Qwen2.5-VL-7B-Instruct2026.0352.7
GSPOBackbone=Qwen2.5-VL-7B-Instruct2026.0352.1
BaselineBackbone=Qwen2.5-VL-7B-Instruct2026.0349.5