Language Modeling on Wikitext-2 word-level (dev)
66.9PPLAWD-LSTM + Phrase Induction + finetuning
Evaluation results
| Method | Links | |
|---|---|---|
| AWD-LSTM + Phrase Induction + finetuning#Params=33M2019.06 | 66.9 | |
| AWD-LSTM + Phrase Induction#Params=33M2019.06 | 68.4 | |
| AWD-LSTM + finetuning#Params=33M2019.06 | 68.6 | |
| 2-Layer Skip Conn. LSTM#Params=24M2019.06 | 69.1 | |
| 1-Layer LSTM#Params=24M2019.06 | 69.3 | |
| VLSTM + augmented loss#Params=28M2019.06 | 91.5 | |
| Variational LSTM (tied)#Params=28M2019.06 | 92.3 |