List your company

Mathematical Reasoning on Hendrycks MATH (GRPO Convergence Metrics)

36.7Peak AccuracyGXPO
33.99634.69835.436.102May 7, 2026

Evaluation results

MethodLinks
GXPOModel=Llama3.2-3B, k=102026.0536.76046.062.331801.33
GXPOModel=Llama3.2-3B, k=32026.0536.19653.696.722.11951.23
SFPOModel=Llama3.2-3B, k=102026.0535.2580318.680.768700.28
SFPOModel=Llama3.2-3B, k=32026.0534.658039.071.563200.75
GRPOModel=Llama3.2-3B, k=–2026.0534.1240114.1412401