WMT
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
WMT24
86.16GRF
21
WMT 23
89.79GRF
21
WMT newstest 2010 (test)
28.29BLEU
21
WMT'16 Romanian-English (Ro-En) (test)
40.4BLEU
21
WMT Chinese-English 2017 (test)
27.4BLEU
21
WMT Xx ⇒ En 24++
85sCMT Score
20
WMT ru 19
33.8PRR
20
WMT de 19
46.5PRR
20
WMT fr 14
39.1Prediction Ranking Rate
20
WMT de 14
34.8Prediction Ranking Rate
20
WMT FI-EN 19
0.863COMET
20
WMT En-De 14 (val)
27.61BLEU
20
WMT De-En 23
29.4BLEU
20
WMT MQM Segment-level 22
60.1Score (En-De)
19
WMT segment-level 2019 (test)
44.5Pearson R
19
WMT19 English-German (En-De) (test)
46.03BLEU
19
WMT (newstest2014)
35BLEU
19
WMT document-level Medium Resource X→En 24++
84.97COMET-22 Score
18
WMT (document-level) Medium Resource En→X 24++
87.39COMET-22
18
WMT (document-level) High Resource En→X 24++
85.35COMET-22
18
WMT document-level 24++
32.45High Resource MT Score (En->X)
18
WMT15 En-Fr nt2014 OOD ID (test)
89.91AUROC
18
ndt2015 OOD WMT15 En-Fr ID (test)
0.919AUROC
18
WMT15 Subtitles OOD En-Fr ID (test)
95.41AUROC
18
WMT15 En-Fr (ID) IT (OOD) (test)
97.41AUROC
18