Machine Translation on IWSLT En-Fr TED 2013 (test)
45.5BLEU ScoreOur Data Diversification with Transformer
Evaluation results
| Method | Links | |
|---|---|---|
| Our Data Diversification with TransformerModel Architecture=Transformer, Training Strategy=Data Diversification2019.11 | 45.5 | |
| Our Data Diversification with Dynamic ConvModel Architecture=Dynamic Conv, Training Strategy=Data Diversification2019.11 | 45.2 | |
| Transformer with Ens-DistillModel Architecture=Transformer, Training Strategy=Ensemble Distillation2019.11 | 44.3 | |
| Transformer with Distill (T=S)Model Architecture=Transformer, Training Strategy=Knowledge Distillation (T=S)2019.11 | 44.1 | |
| TransformerModel Architecture=Transformer, Training Strategy=Baseline2019.11 | 44 | |
| Dynamic ConvModel Architecture=Dynamic Conv, Training Strategy=Baseline2019.11 | 43.8 | |
| Transformer with Distill (T>S)Model Architecture=Transformer, Training Strategy=Knowledge Distillation (T>S)2019.11 | 43.4 |