Machine Translation on WMT'14 De-En (newstest2014)
34.94BLEUBiBERT Contextualized Embeddings + Stochastic Layer Selection
Evaluation results
| Method | Links | |
|---|---|---|
| BiBERT Contextualized Embeddings + Stochastic Layer SelectionStochastic Layer Selection=true2021.09 | 34.94 | |
| BiBERT Contextualized Embeddings + Stochastic Layer Selection + Fine-TuningStochastic Layer Selection=true, Fine-tuning Protocol=Fine-Tuning2021.09 | 34.68 | |
| BiBERT Contextualized Embeddings + Stochastic Layer Selection + Dual-Directional TrainingStochastic Layer Selection=true, Training Strategy=Dual-Directional Training2021.09 | 34.54 | |
| BERT Initialization (12 layers)Model Architecture (Layers)=12 layers2021.09 | 33.6 |