WMT
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
WMT'14 1.2.10 (test)
46.4BLEU
7
WMT fr-en
26.13BLEU
7
WMT en-fr
27.97BLEU Score
7
WMT 19
99BLEU
6
WMT EN-FR_FR 24++
0.652SLI
6
WMT all 54 LPs 24++
0.615SLI #1
6
WMT14 1.1:1
3.32Cohen's d
6
WMT German-English 16
31.789BLEU
6
WMT MQM (ZH-EN) 2023 (test)
50.25Precision
6
WMT MQM EN-DE 2023 (test)
39.02Precision
6
WMT MQM (EN-DE) 2022 (test)
42.66Precision
6
WMT16
35.81BLEU Score (Linewise)
6
WMT en-is 24
0.36Kendall Correlation
6
WMT en-cs 24
0.318Kendall Correlation
6
WMT 24 cs-uk
0.825Kendall Correlation
6
WMT xh-zu 21
0.323Kendall Correlation
6
WMT en-ha 21
0.386Kendall Correlation
6
WMT 24
0.945Quality (cs-uk)
6
WMT 21
54.3Score (en-ha)
6
WMT24++ eng → 22 (<200)
32.2spBLEU
6
WMT eng → 22 24++
34.3spBLEU
6
WMT Ro→En (+BT) 2016 (test)
39.1BLEU
6
WMT Ru-En 20
40.48BLEU
6
WMT De-En 20
44.09BLEU
6
WMT De-En 19
87.5COMET
6