List your company

Mathematical Reasoning on GSM8K (test) (EM, ΔGSM8K)

66.26GSM8K EMEPnG
46.281651.468356.65561.8417Jun 30, 2026Jul 1, 2026Jul 2, 2026

Evaluation results

MethodLinks
EPnGBackbone=OLMoE, Params. (%)=0.722026.0766.26—38.43
ESFTBackbone=OLMoE, Params. (%)=5.082026.0765.13—38.06
FFTBackbone=OLMoE, Params. (%)=1002026.0764.97—38.54
LoRA (static)Backbone=OLMoE, Params. (%)=0.722026.0764.44—36.7
Starting model2026.0649.7——
FProjBest Ep=1.02026.0648.840.86—
ForaBest Ep=1.02026.0648.840.86—
LwFBest Ep=2.22026.0648.281.42—
Full FTBest Ep=1.02026.0647.432.27—
EWCBest Ep=1.02026.0647.262.44—
L2-SPBest Ep=1.02026.0647.132.57—
WProjBest Ep=1.02026.0647.052.65—