Machine Translation on WMT De⇒En 2021 (test)
32.26BLEUDAT
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DATKnowledge Distillation=w/2024.05 | 32.26 | 60.8 | 83.32 | 71.44 | 92.05 | 13.1 | |
| ATKnowledge Distillation=w/2024.05 | 32.04 | 60.85 | 84.72 | 72.53 | 93.39 | 1 | |
| ATKnowledge Distillation=w/o2024.05 | 31.89 | 60.25 | 84.26 | 71.94 | 92.91 | 1 | |
| DATKnowledge Distillation=w/o2024.05 | 31.69 | 59.6 | 81.12 | 69 | 88.29 | 13.1 | |
| CTCKnowledge Distillation=w/2024.05 | 30.52 | 59.83 | 80.06 | 67.24 | 86.91 | 13.4 | |
| MgMOKnowledge Distillation=w/2024.05 | 30.32 | 59.36 | 81.15 | 67.76 | 89.17 | 13.8 | |
| CMLMKnowledge Distillation=w/2024.05 | 30.25 | 58.4 | 77.14 | 65.63 | 82.98 | 2.4 | |
| CMLMKnowledge Distillation=w/o2024.05 | 29.36 | 57.79 | 76.39 | 65 | 82.07 | 2.4 | |
| MgMOKnowledge Distillation=w/o2024.05 | 28.89 | 58.11 | 77.76 | 64.08 | 83.51 | 13.8 | |
| NATKnowledge Distillation=w/2024.05 | 27.55 | 56.56 | 75.5 | 62.72 | 76.89 | 15.2 | |
| CTCKnowledge Distillation=w/o2024.05 | 27.35 | 56.53 | 75.38 | 61.79 | 79.14 | 13.5 | |
| NATKnowledge Distillation=w/o2024.05 | 16.85 | 43.46 | 55.87 | 44.8 | 36.94 | 15.5 |