List your company

Mathematical Reasoning on GSM8K (Acc, SL, SE)

100AccuracyVanilla
82.663287.164191.66596.1659Mar 23, 2026

Evaluation results

MethodLinks
Vanilla2026.031002,0434.89
L12026.031001,2597.94
EAT2026.031001,5946.27
ROM_CSCCounterfactual Self-Correction=true2026.031001,0429.6
ROM2026.0399.21,2547.91
Cut2048max tokens=20482026.0398.332,0414.82
Cut1024max tokens=10242026.0396.672,0414.74
Cut512max tokens=5122026.0383.332,0414.08