List your company

Mathematical Reasoning on Hard problems

1,550Pass@1POPE (+ hard + guide)
-2.72400.39803.51,206.61Jan 26, 2026

Evaluation results

MethodLinks
POPE (+ hard + guide)Training Mixture=Hard problems only2026.011,55042.53
POPE (+ hard + guide + easy)Training Mixture=Hard (256) + Easy (256)2026.011,43238.93
POPE (+ hard + guide + 1K easy problems)Training Mixture=Hard (256) + Easy (1K)2026.011,39836.42
SFT (+ hard)Training Mixture=Hard problems only2026.011,35532.89
SFT (+ hard + easy)Training Mixture=Hard (256) + Easy (256)2026.0182223.81
Prefix + rejection-sampled SFTTraining Mixture=Hard problems only2026.0151424.5
SFT (+ hard + 1K easy problems)Training Mixture=Hard (256) + Easy (1K)2026.0122425.12
Full-oracle SFTTraining Mixture=Hard problems only2026.0120012.37
Qwen3-4B-InstructType=Base model2026.01577.42