Data-to-text generation on WebNLG en
55.52ROUGE-2PEGASUS-2B (calibrated)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PEGASUS-2B (calibrated)#params=2B, beam size=10, training=SLiC2022.09 | 55.52 | 78.09 | 65.06 | |
| ULL#params=20B2022.09 | 55.4 | — | — | |
| Bakshi et al.Evaluation Protocol=Finetuning2022.04 | 53.5 | — | — | |
| PEGASUS-2B (fine-tuned)#params=2B, beam size=52022.09 | 52.97 | 76.96 | 62.56 | |
| PaLMModel Scale=540B, Evaluation Protocol=Finetuning2022.04 | 49.3 | — | — | |
| PaLMModel Scale=62B, Evaluation Protocol=Finetuning2022.04 | 48.6 | — | — | |
| PaLMModel Scale=8B, Evaluation Protocol=Finetuning2022.04 | 47.6 | — | — | |
| PaLMModel Scale=540B, Evaluation Protocol=1-shot2022.04 | 44.4 | — | — | |
| T5Model Scale=XXL, Evaluation Protocol=Finetuning2022.04 | 39.6 | — | — | |
| PaLMModel Scale=62B, Evaluation Protocol=1-shot2022.04 | 38.6 | — | — | |
| LaMDAModel Scale=137B, Evaluation Protocol=1-shot2022.04 | 30.5 | — | — | |
| PaLMModel Scale=8B, Evaluation Protocol=1-shot2022.04 | 29.1 | — | — |