List your company

Mathematical Problem Solving on MATH 500 (test)

85.4Pass@1 AccPRM-CoT (Process-Aware)
63.66469.30774.9580.593Dec 2, 2025Jan 3, 2026Feb 4, 2026Mar 8, 2026Apr 9, 2026May 11, 2026Jun 13, 2026

Evaluation results

MethodLinks
PRM-CoT (Process-Aware)Base Policy=Qwen2.5-Math-7B2025.1285.483.69192.2———————
RLVR (Ground Truth)Base Policy=Qwen2.5-Math-7B2025.1283.782.59192———————
PRM (Process-Aware)Base Policy=Qwen2.5-Math-7B2025.1282.881.4390.891.4———————
SFT (Baseline)Base Policy=Qwen2.5-Math-7B2025.127978.284.888.6———————
GRPOBase Model=Qwen2.5-Math-7B, Temperature=0.6, top-p=0.95, Max tokens=4096, Scoring=exact any-boxed answer scoring2026.0677.5—86.688.381.784.689.791929387.2
BBGBase Model=Qwen2.5-Math-7B, Temperature=0.6, top-p=0.95, Max tokens=4096, Scoring=exact any-boxed answer scoring2026.0675.6—89.491.681.686.393.494.895.896.689.5
REINFORCEBase Model=Qwen2.5-Math-7B, Temperature=0.6, top-p=0.95, Max tokens=4096, Scoring=exact any-boxed answer scoring2026.0675.5—85.987.979.782.788.889.5919285.9
W-RFBase Model=Qwen2.5-Math-7B, Temperature=0.6, top-p=0.95, Max tokens=4096, Scoring=exact any-boxed answer scoring2026.0675.5—8788.980.884.590.792.293.293.887.4
Base ModelBase Model=Qwen2.5-Math-7B, Temperature=0.6, top-p=0.95, Max tokens=4096, Scoring=exact any-boxed answer scoring2026.0664.5—88.290.876.283.892.69495.396.286.8