Machine Translation on WMT En-Fr (newstest2014)
37.5BLEUMASS
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MASSArchitecture=6-layer Transformer2019.05 | 37.5 | — | |
| XLMArchitecture=6-layer Transformer2019.05 | 33.4 | — | |
| Lample et al. (2018)Architecture=4-layer Transformer2019.05 | 25.14 | — | |
| Yang et al. (2018)Architecture=4-layer Transformer2019.05 | 16.97 | — | |
| Artetxe et al. (2017)Architecture=2-layer RNN2019.05 | 15.13 | — | |
| Lample et al. (2017)Architecture=3-layer RNN2019.05 | 15.05 | — | |
| NMT after RLArchitecture=GNMT, Ensemble=8 models, Training Mode=RL-refined2016.09 | 0.4116 | 4.44 | |
| NMT before RLArchitecture=GNMT, Ensemble=8 models, Training Mode=ML-trained2016.09 | 0.4035 | 4.46 | |
| PBMTSystem Type=Phrase-based statistical machine translation2016.09 | 0.37 | 3.87 | |
| HumanType=Human reference translations2016.09 | — | 4.82 |