FLORES
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Flores-200 (test)
88.2COMET
8
FLORES Zh-En
26.25BLEU
8
FLORES-200 NLLB 1.0
95.4F1 Score
8
FLORES-200 OpenLID 1.0
94.7F1 Score
8
FLORES-200 FT176 1.0
99.1F1 Score
8
FLORES-200 CLD3 1.0
98.3F1 Score
8
FLORES-AE33 Type III
0.81MAE
8
Flores-101 Averaged over all language pairs (test)
0.508COMET
8
FLORES en->xx (test)
47.8Score (en->de)
8
FLORES drop perturbation (test)
-68.01Average Error
8
FLORES swap perturbation (test)
-72.14Average Score
8
FLORES-200 10 low-resourced languages
21.2Performance Score (kaz_Cyrl)
7
FLORES-200 199 languages (test)
97.16F1
7
Flores (test)
17.5BLEU
7
Flores
44.2ChrF++ Score
7
FLORES-200 snd_Arab-English (test)
87.5COMET-22
7
FLORES-200 npi_Deva-English (test)
91.1COMET-22
7
FLORES-200 khm_Khmr-English (test)
87.1COMET-22
7
FLORES-200 amh_Ethi-English (test)
88COMET-22
7
FLORES+ 30-lang
0.027Compression Rate
6
Flores-200 zh-bo
21.56BLEU
6
Flores 66 languages
43.5ChrF
6
FLORES-200 cmn → 27
28.3spBLEU
6
FLORES-200 kor → 27
28.3spBLEU
6
FLORES-200 jpn → 27
27.3spBLEU
6