WMT
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
WMT English-French '14
45.6BLEU
14
WMT En→De 23
57.43BLRT
14
WMT 25
83.7Score
14
WMT 24 official EU languages 25
83.9Score
14
WMT 24 official EU languages 24++
84Score
14
WMT non-EU languages 25 (test)
83.7Score
14
WMT X→EN 2022 (test)
85.93Czech Score
14
WMT is-en
37.5BLEU
14
WMT16
32.31BLEU
14
WMT EN-DE 2017 (test)
33.7BLEU
14
WMT En-Fi 17 (test)
30.1BLEU (tokenized)
14
WMT MQM 2025
37.34Precision
13
WMT literary translation (zh→de) 24
94.54SEGALE-COMET Score
13
WMT literary translation (zh→ru) 24
93.74SEGALE_comet
13
WMT literary translation (zh→en) 24
94.62SEGALE Comet Score
13
WMT EN–FI 24
59.44chrF++
13
WMT22 Sah-Ru (test)
59.5COMET Score
13
WMT Reverse (R) 2014-2018 (test)
40.2BLEU
13
WMT en-de 16 (test)
41.2BLEU
13
WMT En-De 2013
28.72BLEU
13
WMT German-English (test)
41BLEU
13
WMT
3.2PPL
13
WMT newstest 2014 (test)
44.3Detokenized BLEU
13
WMT zh→en 2023 (test)
81.3AUC
12
WMT document-level Low Resource X→Zh 24++
83.8COMET-22 Score
12