List your company

Math Reasoning on Open-RS

98.47Relative Overall ScoreJensen-Shannon
-3.938822.648149.23575.8219Feb 5, 2026

Evaluation results

MethodLinks
Jensen-ShannonBackbone=Qwen2.5-Math-7B2026.0298.472
Reverse KLBackbone=Qwen2.5-Math-7B2026.0296.782.4
KLBackbone=Qwen2.5-Math-7B2026.0295.163
Jensen-ShannonBackbone=Qwen2.5-Math-1.5B2026.0293.32.4
PearsonBackbone=Qwen2.5-Math-7B2026.0291.865
HellingerBackbone=Qwen2.5-Math-7B2026.0291.74.4
GRPOBackbone=Qwen2.5-Math-7B2026.0290.525.8
Total VariationBackbone=Qwen2.5-Math-1.5B2026.0286.934
Total VariationBackbone=Qwen2.5-Math-7B2026.0286.054.8
KLBackbone=Qwen2.5-Math-1.5B2026.0285.664
PearsonBackbone=Qwen2.5-Math-1.5B2026.0285.193.6
GRPOBackbone=Qwen2.5-Math-1.5B2026.0282.444.6
Reverse KLBackbone=Qwen2.5-Math-1.5B2026.0281.655.2
HellingerBackbone=Qwen2.5-Math-1.5B2026.0278.473.6
BaseBackbone=Qwen2.5-Math-1.5B2026.0208
BaseBackbone=Qwen2.5-Math-7B2026.0208