WMT
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
WMT de-en & fr-en 14 (test)
0.92Score
26
WMT ZH-EN 22
82.7COMET
26
WMT Original 2014-2018 (test)
37.8BLEU
26
WMT En-De 2016
41.2BLEU
26
WMT document-level En→Ja 24 (test)
4.373M-Prometheus Score
25
WMT Ja→En 21 (test)
4.413M-Prometheus Score
25
WMT ZH-EN 24
81.06COMETkiwi
25
WMT Zh-En (test)
25.55BLEU Score
25
WMT En-De 2022 (test)
87.4COMET
25
WMT 2019 (test)
0.327de-en
25
WMT24++ es
29.33BLEU
24
WMT24++ ru
14.54BLEU
24
WMT fr 24++
24.74BLEU
24
WMT de 24++
20.83BLEU
24
WMT xx→en 24
89.97XCOMET
24
WMT en→xx 24
87.6XCOMET
24
WMT 2023 (test)
88.5COMET
24
WMT14 DE-EN
32.09SacreBLEU
24
WMT en-de
30.17BLEU
24
WMT EN→ZH 23
81.11KIWI Score
23
WMT 22 (test)
33.5BLEU
23
WMT en→ru (test)
30.96BLEU
23
WMT En-Zh 2023 (test)
80.7KIWI Score
22
WMT En-De 14
27.78SacreBLEU
22
WMT19 Zh-En (test)
27.25BLEU
22