List your company

Mathematical Reasoning on PolyMath

20.9Accuracyself-cons (ours)
5.39.3513.417.45May 31, 2026

Evaluation results

MethodLinks
self-cons (ours)Model Size=14B2026.0520.9
self-cons (ours)Model Size=7B2026.0520.5
S1Model Size=7B2026.0520.2
S1Model Size=14B2026.0519.5
baseModel Size=7B2026.0518
baseModel Size=14B2026.0517.5
LIDRModel Size=7B2026.0516.2
LIDRModel Size=14B2026.0515.2
self-cons (ours)Model Size=1.5B2026.058.1
S1Model Size=1.5B2026.058
baseModel Size=1.5B2026.057.5
LIDRModel Size=1.5B2026.055.9