List your company

Mathematical Reasoning on MATH 500 (Pass@32)

0.941Pass@32OPEFO
0.919160.924830.93050.93617May 12, 2026

Evaluation results

MethodLinks
OPEFOBackbone=Qwen2.5-Math-7B2026.050.941
KL-CovBackbone=Qwen2.5-Math-7B2026.050.937
Clip-CovBackbone=Qwen2.5-Math-7B2026.050.935
GRPO (Strict)Backbone=Qwen2.5-Math-7B2026.050.933
Clip-higherBackbone=Qwen2.5-Math-7B2026.050.927
GRPOBackbone=Qwen2.5-Math-7B2026.050.925
Entropy-RegBackbone=Qwen2.5-Math-7B2026.050.92