Natural Language Inference on CB SuperGLUE (test dev)
84AccuracyAdam
Evaluation results
| Method | Links | |
|---|---|---|
| AdamModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 84 | |
| SubZeroModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 67.8 | |
| ZO-MuonModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 64.3 | |
| MeZOModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 60.7 | |
| HiZOOModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 60.7 | |
| LOZOModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 57.1 | |
| LoRAModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 55.4 | |
| S-MeZOModel=Gemma2-2B, Evaluation Protocol=Fine-tuning2026.02 | 53.6 |