Question Answering on BoolQ SuperGLUE (test dev)
90AccuracyAdam
Evaluation results
| Method | Links | |
|---|---|---|
| AdamModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 90 | |
| LoRAModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 75.2 | |
| ZO-MuonModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 74.6 | |
| HiZOOModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 66.6 | |
| MeZOModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 66.1 | |
| LOZOModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 65 | |
| S-MeZOModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 63.9 | |
| SubZeroModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 63.5 |