SMT + NMT (tuning and joint refinement)

Reported on 6 benchmarks across 1 task · 1 paper · 6 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Natural Language Processing6 results

Machine TranslationonWMT2014 English-French
BLEU· 2019-02-04
36.2
best: 38.27 (BERT-fused NMT)
SOTA
An Effective Approach to Unsupervised Machine Translation arXiv:1902.01313
Machine TranslationonWMT2014 French-English
BLEU· 2019-02-04
33.5
best: 39.2 (GPT-3 175B (Few-Shot))
SOTA
An Effective Approach to Unsupervised Machine Translation arXiv:1902.01313
Machine TranslationonWMT2014 English-German
BLEU· 2019-02-04
22.5
SOTA
An Effective Approach to Unsupervised Machine Translation arXiv:1902.01313
Machine TranslationonWMT2016 English-German
BLEU· 2019-02-04
26.9
best: 29.7 (GPT-3 175B (Few-Shot))
SOTA
An Effective Approach to Unsupervised Machine Translation arXiv:1902.01313
Machine TranslationonWMT2014 German-English
BLEU· 2019-02-04
27
SOTA
An Effective Approach to Unsupervised Machine Translation arXiv:1902.01313
Machine TranslationonWMT2016 German-English
BLEU· 2019-02-04
34.4
best: 40.6 (GPT-3 175B (Few-Shot))
SOTA
An Effective Approach to Unsupervised Machine Translation arXiv:1902.01313