List your company

Mathematical Reasoning on GSM8K In-Domain (Accuracy)

65.88Accuracy (GSM8K ID)AdamW
23.853634.764345.67556.5857May 12, 2026

Evaluation results

MethodLinks
AdamWBase Model=Qwen2.5-1.5B2026.0565.88
PionBase Model=Qwen2.5-1.5B2026.0565.76
MuonBase Model=Qwen2.5-1.5B2026.0565.27
AdamWBase Model=Llama3.2-3B2026.0559.87
BaseBase Model=Qwen2.5-1.5B2026.0559.81
PionBase Model=Llama3.2-3B2026.0558.83
MuonBase Model=Llama3.2-3B2026.0557.77
BaseBase Model=Llama3.2-3B2026.0525.47