List your company

STEM Reasoning on Minerva (Avg@32, Pass@16)

56.57Avg@32Training-time reweighting
16.821227.140637.4647.7794Mar 23, 2026

Evaluation results

MethodLinks
Training-time reweightingBackbone=Qwen3-8B-Base2026.0356.5776.78
DAPOBackbone=Qwen3-8B-Base2026.0355.0476.98
Training-time reweightingBackbone=Qwen2.5-Math-7B2026.0349.7270.37
PPLBackbone=Qwen2.5-Math-7B2026.0348.6868.69
DominateBackbone=Qwen2.5-Math-7B2026.0347.0164.59
DAPOBackbone=Qwen2.5-Math-7B2026.0346.4369.44
BaseBackbone=Qwen3-8B-Base2026.0329.870.43
BaseBackbone=Qwen2.5-Math-7B2026.0318.3561.04