WMT
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
WMT en→xx (test)
32.17BLEU
10
WMT xx-en Average
37.36BLEU
10
WMT-6 base (test)
20.48BLEU
10
WMT-6 base (test)
22.39BLEU
10
WMT Out-of-distribution from PAWS
76.4Macro F1
10
WMT Out-of-distribution from QQP
75.5Macro F1
10
WMT De-En (test)
32.78BLEU
10
WMT En-De (S) 19 (test)
36.9BLEU
10
WMT En-Es 13 (test)
35.4Tokenized BLEU
10
WMT EN-DE 2014 (test)
28.3BLEU Score
10
WMT14 English-to-German big (test)
30.1BLEU
10
WMT English-German 2014 (newstest)
35BLEU (tok)
10
WMT English-German 2015 (newstest)
26.5BLEU
10
WMT25 Average (test)
63Group-by-Item Accuracy (Tie Calibrated)
9
WMT25 English → Ukrainian (test)
63Group-by-Item Accuracy (with tie calibration)
9
WMT25 Czech → German (test)
63Group-by-Item Accuracy (with tie calibration)
9
WMT English → Ukrainian 2025 (test)
85Soft Pairwise Accuracy
9
WMT Czech → German 2025 (test)
91Soft Pairwise Accuracy
9
WMT EN→RO 16
32.86SacreBLEU
9
WMT 25 (ESA)
17Precision
9
WMT 17 (val)
70.921Accuracy
9
WMT De-En 2022
33.87BLEU
9
WMT24++ (ko,de,es,ja,vi)
96.7Score
9
WMT22 Liv-En news
47.1COMET Score
9
WMT22 En-Liv news
56.7COMET
9