WMT
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
WMT Pl-En 2020 (test)
23.18BLEU
6
WMT (test)
43BLEU (De)
6
WMT En-Ro
35.28BLEU
6
WMT English-German (en-de) 19
44.9BLEU
6
WMT En-Tr 17 (test)
21.4BLEU (tokenized)
6
WMT All resource levels
23.5BLEU
6
WMT Rich resource, >10M
29BLEU
6
WMT Medium resource, 1M~10M
22.2BLEU
6
WMT Low resource, <1M
18.8BLEU
6
WMT EN-RO 2016 (dev)
34.74BLEU
6
WMT DE-EN 2014 (dev)
31.66BLEU
6
WMT EN-DE 2014 (dev)
26.13BLEU Score
6
WMT News 2017 (test)
90.5BLEU-2
6
WMT En-De 2018 (test)
43.8SacreBLEU
6
WMT French-English 2014 (test)
39.2BLEU
6
WMT English-German 2014 (newstest13)
27.2BLEU
6
WMT English-German (EnDe) 2014 1.0 (test)
29.3BLEU
6
WMT de-en 2016 (newstest)
35.2BLEU
6
WMT fr-en 2014 (newstest)
34.9BLEU
6
WMT English-German '14 (val)
3.69Perplexity
6
WMT 2014 (newstest14)
26.1BLEU
6
WMT English-to-French No UNK 2014 (test)
36.15BLEU
6
WMT En-Zh 2020 (test)
44.3SacreBLEU
6
WMT English-Chinese 2024
99.3ConsistencyChecker
6
WMT Czech-Ukrainian 2024
98.1ConsistencyChecker
6