List your company

Mathematical Reasoning on Math Out-of-Domain

67.59AccuracyBase
60.15462.084564.01565.9455May 12, 2026

Evaluation results

MethodLinks
BaseBase Model=Llama3.2-3B2026.0567.59
BaseBase Model=Qwen2.5-1.5B2026.0564.83
PionBase Model=Qwen2.5-1.5B2026.0562.16
AdamWBase Model=Qwen2.5-1.5B2026.0562.13
MuonBase Model=Qwen2.5-1.5B2026.0561.22
MuonBase Model=Llama3.2-3B2026.0561.2
AdamWBase Model=Llama3.2-3B2026.0560.86
PionBase Model=Llama3.2-3B2026.0560.44