List your company

Mathematical Reasoning on MultiArith (Acc, RL, SL, RE, SE)

100AccuracyVanilla
97.498.07598.7599.425Mar 23, 2026

Evaluation results

MethodLinks
Vanilla2026.031009971,18110.038.47
EAT2026.031009581,14210.448.76
Vanilla2026.03100—1,181—8.47
EAT2026.03100—1,142—8.76
ROMCSC2026.0399.246961621.1516.1
ROM_CSCCounterfactual Self-Correction=true2026.0399.2—616—16.1
Cut5122026.0398.335121,17919.218.34
Cut512max tokens=5122026.0398.33—1,179—8.34
ROM2026.0398.361376816.0412.8
ROM2026.0398.3—768—12.8
L12026.0397.575893012.8610.48
Cut20482026.0397.59341,17910.448.27
Cut10242026.0397.58241,17911.838.27
L12026.0397.5—930—10.48
Cut2048max tokens=20482026.0397.5—1,179—8.27
Cut1024max tokens=10242026.0397.5—1,179—8.27