Machine Translation on En-De 2016 (test)
43.6BLEUmBART + MT* w/ adapters
Evaluation results
| Method | Links | ||
|---|---|---|---|
| mBART + MT* w/ adaptersObjectives=NMT + MLM, trainable_params=12.6M2022.12 | 43.6 | 0.697 | |
| VGAMTObjectives=MMT + VMLM, trainable_params=13.2M2022.12 | 43.3 | 0.694 | |
| TLM + MT*Objectives=NMT, trainable_params=42M2022.12 | 40 | 0.457 | |
| VTLM + MMT*Objectives=MMT, trainable_params=44M2022.12 | 39.4 | 0.439 | |
| Gated Fusion*Objectives=MMT, trainable_params=2.8M2022.12 | 38.7 | 0.378 | |
| Graph-MMT*Objectives=MMT, trainable_params=4.1M2022.12 | 38.6 | 0.368 | |
| Vanilla MT*Objectives=NMT, trainable_params=4.1M2022.12 | 38.5 | 0.394 | |
| mBART + MT*Objectives=NMT, trainable_params=-2022.12 | 36.2 | 0.595 |