List your company

Mathematical Problem Solving on MATH-HARD (Accuracy)

77.3Accuracy (MATH-HARD)KACE
69.70871.67973.6575.621May 30, 2026

Evaluation results

MethodLinks
KACESolver model=gpt-4o-mini2026.0577.3
Tiered Best-of-NSolver model=gpt-4o-mini2026.0573.3
Best-of-5 self-consistencySolver model=gpt-4o-mini2026.0572.7
Tiered + GEPASolver model=gpt-4o-mini2026.0572
Tiered + ACESolver model=gpt-4o-mini2026.0570