Natural Language Generation on NLG 3 tasks one-shot
24.7Average Exact MatchLion
Evaluation results
| Method | Links | |
|---|---|---|
| LionModel Size=7.5B, Optimizer=Lion, #Tokens=300B, Evaluation Protocol=One-shot2023.02 | 24.7 | |
| AdafactorModel Size=7.5B, Optimizer=Adafactor, #Tokens=300B, Evaluation Protocol=One-shot2023.02 | 24.1 | |
| PaLMModel Size=8B, Optimizer=not specified, #Tokens=780B, Evaluation Protocol=One-shot2023.02 | 23.9 | |
| GPT-3Model Size=6.7B, Optimizer=AdamW (implied), #Tokens=300B, Evaluation Protocol=One-shot2023.02 | 23.1 | |
| LionModel Size=2.1B, Optimizer=Lion, #Tokens=300B, Evaluation Protocol=One-shot2023.02 | 16.5 | |
| AdafactorModel Size=2.1B, Optimizer=Adafactor, #Tokens=300B, Evaluation Protocol=One-shot2023.02 | 15.6 | |
| LionModel Size=1.1B, Optimizer=Lion, #Tokens=300B, Evaluation Protocol=One-shot2023.02 | 12.1 | |
| AdafactorModel Size=1.1B, Optimizer=Adafactor, #Tokens=300B, Evaluation Protocol=One-shot2023.02 | 11.1 |