List your company

Mathematical Problem Solving on AMC'23 (test)

66.3Pass@1 AccPRM-CoT (Process-Aware)
3.48419.79236.152.408Oct 28, 2025Nov 2, 2025Nov 8, 2025Nov 14, 2025Nov 20, 2025Nov 26, 2025Dec 2, 2025

Evaluation results

MethodLinks
PRM-CoT (Process-Aware)Base Policy=Qwen2.5-Math-7B2025.1266.362.1775.379.8——
PRM (Process-Aware)Base Policy=Qwen2.5-Math-7B2025.1262.558.7571.0875.9——
RLVR (Ground Truth)Base Policy=Qwen2.5-Math-7B2025.1259.156.369.8874.7——
SFT (Baseline)Base Policy=Qwen2.5-Math-7B2025.1252.54857.8366.27——
DAPOsampling_strategy=LATR2025.1045.3—65—8831,920
DAPOsampling_strategy=Stochastic Sampling2025.1037.8—62.7—1,0752,116
GRPOsampling_strategy=LATR2025.1035.6—60.3—8381,537
GRPOsampling_strategy=Stochastic Sampling2025.1032.8—59.7—8331,622
Qwen2.5-3Bbaseline=True2025.105.9—20.7—9632,255