Mathematical Reasoning on GSM8K In-Domain (Accuracy)
65.88Accuracy (GSM8K ID)AdamW
Evaluation Results
| Method | Links | |
|---|---|---|
| AdamWBase Model=Qwen2.5-1.5B2026.05 | 65.88 | |
| PionBase Model=Qwen2.5-1.5B2026.05 | 65.76 | |
| MuonBase Model=Qwen2.5-1.5B2026.05 | 65.27 | |
| AdamWBase Model=Llama3.2-3B2026.05 | 59.87 | |
| BaseBase Model=Qwen2.5-1.5B2026.05 | 59.81 | |
| PionBase Model=Llama3.2-3B2026.05 | 58.83 | |
| MuonBase Model=Llama3.2-3B2026.05 | 57.77 | |
| BaseBase Model=Llama3.2-3B2026.05 | 25.47 |