List your company

Mathematical Reasoning on Minerva (Pass@1, Pass@32)

27.3Pass@1DynaMO
14.317.67521.0524.425Feb 22, 2026

Evaluation results

MethodLinks
DynaMOBackbone=Qwen2.5-Math-7B2026.0227.347.2
W-REINFORCEBackbone=Qwen2.5-Math-7B2026.0226.449.5
Fork TokensBackbone=Qwen2.5-Math-7B2026.0226.142.4
Clip-COVBackbone=Qwen2.5-Math-7B2026.0225.445.9
KL-COVBackbone=Qwen2.5-Math-7B2026.0224.539.9
Clip-HigherBackbone=Qwen2.5-Math-7B2026.022446.1
Entropy AdvantagesBackbone=Qwen2.5-Math-7B2026.0223.743.7
Entropy LossBackbone=Qwen2.5-Math-7B2026.0223.345.9
GRPOBackbone=Qwen2.5-Math-7B2026.0222.645.4
DynaMOBackbone=Qwen2.5-Math-1.5B2026.0219.444
Entropy AdvantagesBackbone=Qwen2.5-Math-1.5B2026.0218.243
W-REINFORCEBackbone=Qwen2.5-Math-1.5B2026.0218.240.3
Entropy LossBackbone=Qwen2.5-Math-1.5B2026.0217.643.6
GRPOBackbone=Qwen2.5-Math-1.5B2026.0217.242.8
Clip-HigherBackbone=Qwen2.5-Math-1.5B2026.0216.841.3
Fork TokensBackbone=Qwen2.5-Math-1.5B2026.0216.636.8
Clip-COVBackbone=Qwen2.5-Math-1.5B2026.0215.844.3
KL-COVBackbone=Qwen2.5-Math-1.5B2026.0214.840.3