List your company

Mathematical Reasoning on SVAMP original GSM8k-Aug rationale format

64.9Pass@1SFT-CoT
39.73246.26652.859.334May 25, 2026

Evaluation results

MethodLinks
SFT-CoTRationale Format=GSM8k-Aug2026.0564.911.3
Ours-RLRationale Format=GSM8k-Aug2026.0560.85.86
Ours-SFTRationale Format=GSM8k-Aug2026.0558.36.43
CoLaR-2Rationale Format=GSM8k-Aug2026.0554.96.11
CoconutRationale Format=GSM8k-Aug2026.0540.76