lid-neural-608 / eval /comparison.csv
olumideola's picture
benchmark results
dd4186f verified
Raw History Blame Contribute Delete
689 Bytes
,lid-neural-608 (coverage),lid-neural-608 (traffic),lid-lite-608 (coverage),lid-lite-608 (traffic),GlotLID v3,Meta NLLB LID-218
"Held-out test, accuracy",0.867,0.861,0.858,0.843,0.783,0.28
"Held-out test, macro-F1",0.862,0.858,0.856,0.843,0.795,0.182
"Held-out test, 1-5 words",0.666,0.65,0.672,0.656,0.486,0.218
Web-weighted accuracy,0.929,0.96,0.932,0.954,0.946,0.901
"Web-weighted, 1-5 words",0.851,0.906,0.763,0.829,0.808,0.764
Everyday phrases,0.819,0.943,0.762,0.838,0.857,0.867
Tatoeba test,0.93,0.926,0.922,0.914,0.868,0.557
FLORES+ devtest,0.942,0.942,0.934,0.928,0.957,0.874
FLORES+ first 1-5 words,0.715,0.714,0.708,0.708,0.692,0.618
UDHR-LID,0.913,0.914,0.91,0.908,0.908,0.567