List your company

Mathematical Reasoning on BeyondAIME (mean@8)

21Mean@8 ScoreACPO
3.327.9112.517.09Jul 3, 2026

Evaluation results

MethodLinks
ACPOModel=DeepSeek-R1-Distill-Qwen-7B2026.0721
GTPOModel=DeepSeek-R1-Distill-Qwen-7B2026.0719.12
SAPOModel=DeepSeek-R1-Distill-Qwen-7B2026.0718.62
80/20Model=DeepSeek-R1-Distill-Qwen-7B2026.0715.75
ACPOModel=Qwen3-8B-Base2026.0715.5
DAPOModel=DeepSeek-R1-Distill-Qwen-7B2026.0715.12
GRPOModel=DeepSeek-R1-Distill-Qwen-7B2026.0713.62
GTPOModel=Qwen3-8B-Base2026.0712.62
SAPOModel=Qwen3-8B-Base2026.0712.25
80/20Model=Qwen3-8B-Base2026.0712
DAPOModel=Qwen3-8B-Base2026.0711.75
GRPOModel=Qwen3-8B-Base2026.079.5
ACPOModel=Qwen2.5-Math-7B2026.078
GTPOModel=Qwen2.5-Math-7B2026.077.12
SAPOModel=Qwen2.5-Math-7B2026.076.75
80/20Model=Qwen2.5-Math-7B2026.075.25
DAPOModel=Qwen2.5-Math-7B2026.074.5
GRPOModel=Qwen2.5-Math-7B2026.074