Natural Language Processing on 29 public NLP benchmarks (average)
68.1AccuracyGLaM
Evaluation results
| Method | Links | |
|---|---|---|
| GLaMevaluation protocol=Few-shot, FLOPs / token (G)=180, Train energy (MWh)=4562021.12 | 68.1 | |
| GLaMevaluation protocol=One-shot, FLOPs / token (G)=180, Train energy (MWh)=4562021.12 | 65.5 | |
| GPT-3evaluation protocol=Few-shot, FLOPs / token (G)=350, Train energy (MWh)=12872021.12 | 65.2 | |
| GLaMevaluation protocol=Zero-shot, FLOPs / token (G)=180, Train energy (MWh)=4562021.12 | 62.7 | |
| GPT-3evaluation protocol=One-shot, FLOPs / token (G)=350, Train energy (MWh)=12872021.12 | 61.6 | |
| GPT-3evaluation protocol=Zero-shot, FLOPs / token (G)=350, Train energy (MWh)=12872021.12 | 56.9 |