List your company

Mathematical Reasoning on MATH-500 OpenR1 Harder

96.6AccuracyQwen-4B
93.68894.44495.295.956Feb 11, 2026

Evaluation results

MethodLinks
Qwen-4BBackbone=Qwen, Parameters=4B2026.0296.6
RePOBackbone=Qwen, Parameters=4B2026.0296.6
LUFFYBackbone=Qwen, Parameters=4B2026.0293.8