Machine Translation on En-Fr 2017 (test)
61.6BLEUVGAMT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VGAMTObjectives=MMT + VMLM, trainable_params=13.2M2022.12 | 61.6 | 0.921 | |
| mBART + MT* w/ adaptersObjectives=NMT + MLM, trainable_params=12.6M2022.12 | 61.5 | 0.918 | |
| TLM + MT*Objectives=NMT, trainable_params=42M2022.12 | 54.2 | 0.681 | |
| VTLM + MMT*Objectives=MMT, trainable_params=44M2022.12 | 53.6 | 0.672 | |
| Vanilla MT*Objectives=NMT, trainable_params=4.0M2022.12 | 51.6 | 0.568 | |
| Graph-MMT*Objectives=MMT, trainable_params=4.0M2022.12 | 51.5 | 0.589 | |
| Gated Fusion*Objectives=MMT, trainable_params=2.8M2022.12 | 50.8 | 0.58 | |
| mBART + MT*Objectives=NMT, trainable_params=-2022.12 | 48.1 | 0.779 |