List your company

Mathematical Reasoning on ProcessBench GSM8K 1.0 (test)

96AccuracyQwen2.5-Math-PRM
50.7662.50574.2585.995Mar 28, 2025

Evaluation results

MethodLinks
Qwen2.5-Math-PRMModel Size=72B2025.039696.193.898.5
Greedy EDU PRMModel Size=72B2025.0394.29590.999.5
Omega PRMModel Size=72B2025.0390.590.888.293.5
Qwen2.5-Math-PRMModel Size=7B2025.0388.889.583.896
Math-Shepherd PRMModel Size=72B2025.0374.580.367.1100
Sample EDU PRMModel Size=72B2025.037177.863.7100
Omega PRMModel Size=7B2025.0357.53184.419
Math-Shepherd PRMModel Size=7B2025.0357.268.254.591
Greedy EDU PRMModel Size=7B2025.0355.221.886.212.5
Sample EDU PRMModel Size=7B2025.0352.567.751.399.5