Natural Language Understanding on SuperGLUE (test val)
96SST-2 AccuracyAdam
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| AdamModel=LLaMA3-8B2026.05 | 96 | 92 | 86.6 | 90.4 | |
| AdamModel=OPT-13B2026.05 | 95.3 | 80.9 | 83.5 | 89.5 | |
| LoRAModel=LLaMA3-8B2026.05 | 95 | 80.9 | 86.4 | 89.4 | |
| LoRAModel=OPT-13B2026.05 | 94.8 | 78.3 | 80.2 | 88 | |
| ZO-MuonModel=LLaMA3-8B2026.05 | 93.7 | 78 | 82.9 | 88.2 | |
| ZO-MOPIModel=OPT-13B2026.05 | 93.2 | 75.8 | 72 | 83.5 | |
| ZO-MOPIModel=LLaMA3-8B2026.05 | 92.9 | 78 | 82.5 | 88.5 | |
| MeZOModel=LLaMA3-8B2026.05 | 92.7 | 66.8 | 76.7 | 86.7 | |
| ZO-MuonModel=OPT-13B2026.05 | 92.5 | 72.9 | 71.3 | 84.2 | |
| LOZOModel=LLaMA3-8B2026.05 | 92.5 | 66.8 | 79.4 | 89 | |
| LOZOModel=OPT-13B2026.05 | 91.6 | 70.4 | 70 | 84.9 | |
| MeZOModel=OPT-13B2026.05 | 91.4 | 66.1 | 67.3 | 81.8 |