Language Modeling on WSJ PTG (test)
26.8PerplexityGold F-SGS
Evaluation results
| Method | Links | |
|---|---|---|
| Gold F-SGSrepresentation=Fully labeled/anchored2023.05 | 26.8 | |
| Gold B-SGSrepresentation=Binary slices2023.05 | 33.9 | |
| Domain-finetuned (baseline)type=non-symbolic2023.05 | 46.1 | |
| Pretrained GPT-2type=non-symbolic2023.05 | 59.3 |