Data-to-text generation on WTQ (test)
62.25ROUGE-1Fully-supervised fine-tuning
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Fully-supervised fine-tuningTraining setting=Fully-supervised2023.05 | 62.25 | 34.59 | 49.41 | 39.17 | 21.18 | 92.88 | 24.18 | |
| CycleNLGTraining setting=Low-resource cycle training2023.05 | 61.54 | 34.25 | 49.07 | 39.09 | 20.93 | 92.66 | 24.39 | |
| CycleNLGTraining setting=Unsupervised cycle training2023.05 | 61.27 | 33.45 | 48.22 | 39.06 | 20.46 | 92.67 | 23.05 | |
| Low-resource fine-tuningTraining setting=Low-resource2023.05 | 55.89 | 31.6 | 46.73 | 31.98 | 15.34 | 91.91 | 23.36 | |
| Low-resource fine-tuning + additional pre-trainingTraining setting=Low-resource, Additional pre-training=true2023.05 | 55.57 | 30.48 | 44.47 | 33.73 | 15.89 | 91.53 | 22.88 |