WMT
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
WMT DA 2022 (test)
33Completeness Score (AR)
18
WMT DA 2022 (test)
0.16QE Score (ar)
18
WMT 22 (test)
85.05COMET
18
WMT En-Zh 22 (test)
86.8COMET
18
WMT'16 En-Ro (test)
38BLEU
18
WMT News 2020 (test)
0.7Kendall's tau
18
WMT 24++
86.8xCOMET (En->X)
17
WMT EN-UK 2025
0.572Acc*Eq
17
WMT EN-JA 2025
57.3Acc*Eq
17
WMT EN-ZH 2025
56.8Acc*Eq
17
WMT EN-CS 2025
61.4Acc*Eq
17
WMT En–Ja 2023 (dev)
88.2COMET Score
17
WMT 2025 (test)
61.59XCOMET-XXL
17
WMT news Average 2010-2016 (test)
28Average BLEU
17
WMT En-Tr 17
25.8BLEU
17
WMT 14
0.013MAE
16
WMT 24++
10Standard Deviation of Score
16
WMT MQM 2022 (test)
91.6Accuracy (System, 3 LPs)
16
WMT 24++
0.837Score
16
WMT non-EU languages 24++ (test)
83.2Score
16
WMT EN-DE 2022
87.44COMET22
16
WMT Benchmarks Average WMT'14 & WMT'19 (aggregation)
0.29MAE
16
WMT En-De 2019 (val)
0.29MAE
16
WMT En-Fr 2014 (val)
0.28MAE
16
WMT En-De 2014 (val)
0.22MAE
16