Machine Translation on WMT Original 2014-2018 (test)
37.8BLEUFT + Beam BT + TST
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| FT + Beam BT + TSTTraining Strategy=Forward Translation (FT), BT Variant=Beam2023.06 | 37.8 | 64.6 | 61.3 | 76.1 | |
| FT + Beam BT + TSTTraining Strategy=Forward Translation (FT), BT Variant=Beam2023.06 | 37.7 | 64 | 55.4 | 73.8 | |
| FT + Beam BTTraining Strategy=Forward Translation (FT), BT Variant=Beam2023.06 | 37.6 | 63.8 | 54.7 | 73.3 | |
| TST (with Tagged BT)Training Strategy=Bitext, BT Variant=Tagged2023.06 | 37.4 | 63.8 | 57.2 | 74.4 | |
| TST (with Tagged BT)Training Strategy=Bitext, BT Variant=Tagged2023.06 | 37.4 | 64.3 | 62.2 | 76.6 | |
| FT + Beam BTTraining Strategy=Forward Translation (FT), BT Variant=Beam2023.06 | 37.3 | 64.3 | 60.4 | 75.6 | |
| TST (with Noised BT)Training Strategy=Bitext, BT Variant=Noised2023.06 | 37 | 64.1 | 62.3 | 76.5 | |
| Tagged BTTraining Strategy=Bitext, BT Variant=Tagged2023.06 | 37 | 63.9 | 61.6 | 76.2 | |
| Tagged BTTraining Strategy=Bitext, BT Variant=Tagged2023.06 | 36.6 | 63.1 | 56 | 73.9 | |
| Noised BTTraining Strategy=Bitext, BT Variant=Noised2023.06 | 36.6 | 63.6 | 59.8 | 75.7 | |
| TST (with Noised BT)Training Strategy=Bitext, BT Variant=Noised2023.06 | 36.5 | 63.1 | 57.1 | 74.4 | |
| FTTraining Strategy=Forward Translation (FT)2023.06 | 36.4 | 63.1 | 52.4 | 72.5 | |
| Noised BTTraining Strategy=Bitext, BT Variant=Noised2023.06 | 36.2 | 62.6 | 53.4 | 73 | |
| Sampling BTTraining Strategy=Bitext, BT Variant=Sampling2023.06 | 36 | 63 | 61.7 | 76.9 | |
| TST (with Sampling BT)Training Strategy=Bitext, BT Variant=Sampling2023.06 | 35.8 | 63 | 62.5 | 77.2 | |
| TST CascadeTraining Strategy=Bitext, BT Variant=Beam, TST Approach=Cascade2023.06 | 35.3 | 62.8 | 59.3 | 75.8 | |
| BitextTraining Strategy=Bitext2023.06 | 35.2 | 62.1 | 50.1 | 71.8 | |
| FTTraining Strategy=Forward Translation (FT)2023.06 | 33.6 | 61.5 | 56.3 | 74.1 | |
| TST DirectTraining Strategy=Bitext, BT Variant=Beam, TST Approach=Direct2023.06 | 33.3 | 60.8 | 53.3 | 73.9 | |
| TST CascadeTraining Strategy=Bitext, BT Variant=Beam, TST Approach=Cascade2023.06 | 33 | 60.6 | 51.6 | 73.1 | |
| BitextTraining Strategy=Bitext2023.06 | 32.9 | 60.8 | 54.8 | 73.6 | |
| Sampling BTTraining Strategy=Bitext, BT Variant=Sampling2023.06 | 32.7 | 60.2 | 54.5 | 74.2 | |
| TST (with Sampling BT)Training Strategy=Bitext, BT Variant=Sampling2023.06 | 32.6 | 60.3 | 55.9 | 74.7 | |
| Beam BTTraining Strategy=Bitext, BT Variant=Beam2023.06 | 32.1 | 59.2 | 45.9 | 71.7 | |
| TST DirectTraining Strategy=Bitext, BT Variant=Beam, TST Approach=Direct2023.06 | 31.4 | 58.4 | 42.2 | 70.3 | |
| Beam BTTraining Strategy=Bitext, BT Variant=Beam2023.06 | 28.5 | 55 | 28 | 66.2 |