List your company

Mathematical Reasoning on AIME 2025 (Pass@1, TC)

30Pass@1ARPO
5.76812.05918.3524.641Jun 1, 2026

Evaluation results

MethodLinks
ARPOBackbone Model=Qwen2.5-7B-Instruct2026.06301.03
EAPOBackbone Model=Qwen2.5-7B-Instruct2026.0626.60.93
GRPOBackbone Model=Qwen2.5-7B-Instruct2026.0623.30.83
Reinforce++Backbone Model=Qwen2.5-7B-Instruct2026.0623.31.03
AEPOBackbone Model=Qwen2.5-7B-Instruct2026.0623.31.21
ToolStarBackbone Model=Qwen2.5-7B-Instruct2026.06201.4
Reinforce++Backbone Model=Qwen2.5-3B-Instruct2026.0616.71.3
ARPOBackbone Model=Qwen2.5-3B-Instruct2026.0616.71.3
GRPOBackbone Model=Qwen2.5-3B-Instruct2026.0613.31.22
ToolStarBackbone Model=Qwen2.5-3B-Instruct2026.0613.31.11
EAPOBackbone Model=Qwen2.5-3B-Instruct2026.0613.30.89
BaseBackbone Model=Qwen2.5-7B-Instruct2026.0610—
TIRBackbone Model=Qwen2.5-7B-Instruct2026.06100.16
BaseBackbone Model=Qwen2.5-3B-Instruct2026.066.7—
TIRBackbone Model=Qwen2.5-3B-Instruct2026.066.70.2