WMT
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
WMT25 Czech-German (official human annotations)
16.66Precision
8
WMT Bible domain (test)
28.03BLEU
8
WMT en-es 24
0.672Pearson Correlation
8
WMT en-de 24
0.606Pearson Correlation
8
WMT14
25.28BLEU
8
WMT23 En-XX Average (test)
33.9BLEU
8
WMT23 En-Uk (test)
27.4BLEU
8
WMT23 En-Ru (test)
31.5BLEU Score
8
WMT23 En-De (test)
40.9BLEU
8
WMT23 En-Zh (test)
47.5BLEU
8
WMT23 (test)
85.6COMET
8
WMT23 (test)
83.2COMET Score
8
WMT22 newstest en-zh (test)
46.8d-BLEU
8
WMT22 en-de (test)
34.08d-BLEU
8
WMT newstest zh-en WMT22 (test)
29.32d-BLEU
8
WMT22 newstest de-en (test)
38.14d-BLEU
8
WMT
28.6BLEU (en-de)
8
WMT English-Romanian 2016 (test)
32.3BLEU
8
WMT En-Ro (test)
35.4BLEU
8
WMT news 2013 (test)
30.01BLEU
8
WMT newstest 2012
25.98BLEU
8
WMT Machine Translation Using Terminologies Shared Task En-Zh 2021 (test)
87.2Exact Match Acc
8
WMT En-Zh 2022 (test)
50.5SacreBLEU
8
WMT Ja-En 2022 (test)
26.6SacreBLEU
8
WMT Ru-En 2022 (test)
45.7SacreBLEU
8