WMT
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
WMT To-English 2022 (test)
45.1Score (Zh->En)
3
WMT En-Ja 2022 (test)
42.5SacreBLEU
3
WMT En-Ru 2022 (test)
32.7SacreBLEU
3
WMT En-Cs 2022 (test)
41.4SacreBLEU
3
WMT Cs-En 2022 (test)
54.9SacreBLEU
3
WMT De-En 2022 (test)
33.7SacreBLEU
3
WMT Supervised Representative Directions (test)
33.01BLEU (Tokenized Average)
3
WMT Pl->En 2020 (test)
17.1Tokenized BLEU
3
WMT En->Pl 2020 (test)
6.7Tokenized BLEU
3
WMT En-Cs 19
19.9BLEU
3
WMT Lv-En 17
24.4BLEU
3
WMT Fi-En 17
28.5BLEU
3
WMT Et-En 18
28.2BLEU
3
WMT Tr-En 17
23.4BLEU
3
WMT Kk-En 19
12.3BLEU
3
WMT Gu-En 19
60BLEU
3
WMT19 En-Gu
3.2BLEU
3
WMT Russian-English 18 (test)
32.06BLEU
3
WMT Russian-English 18 (dev)
36.38BLEU
3
WMT20
0.2Reward
3
WMT 2022 (test)
0.157BLEU
2
WMT ZH-EN 23
91.05Mean DA
2
WMT EN-ZH 23
85.65Mean DA
2
WMT16 (test)
62.99Recall
2
WMT Instruct 16
36.441BLEU (linewise)
2