Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Languages

Benchmarks

Task NameDataset NameSOTA ResultTrend
Acoustic Discriminability (ABX)5 Languages (sw, ta, th, tr, uk) (dev)
Triphone ABX (WS)3.58
22
Within-Speaker ABX3 languages (test)
Avg ABX (w/o 0h)3.76
7
Natural Language Inference20 languages
Accuracy53.5
6
Named Entity Recognition20 languages
F1 Score66.3
6
Part-of-Speech Tagging6 languages Averaged (test)
F1 Score69.2
4
Sentence Retrieval6 languages Same script split
Accuracy43.9
4
Natural Language Inference6 languages All transfers
Accuracy43
4
Dependency Labeling6 languages All transfers
F1 Score18.1
4
Dependency Labeling6 languages (Different script)
F1 Score15.9
4
Part-of-Speech Tagging6 languages Same script split
F1 Score41.9
4
Cross-language Vocabulary Overlap6 languages All transfers
JSD0.74
4
Cross-language Vocabulary Overlap6 languages (Same script split)
JSD0.62
4
Cross-language Vocabulary Overlap6 languages Different script
JSD0.77
4
Sentence Retrieval20 languages All transfers
Accuracy45.1
3
Sentence Retrieval20 languages Different script
Accuracy44.1
3
Natural Language Inference20 languages Different script
Accuracy37.8
3
Dependency Labeling20 languages Same script
F1 Score19.4
3
Dependency Labeling20 languages Different script
F1 Score16.5
3
Named Entity Recognition20 languages Same script
F1 Score54.3
3
Vocabulary Overlap20 languages Same script
JSD0.58
3
Vocabulary Overlap20 languages Different script
JSD0.75
3
Dependency Labeling20 languages
F1 Score0.545
3
Part-of-Speech Tagging20 languages
F1 Score67.3
3
Masked Language Modeling20 languages
MRR52.6
3
Vocabulary Allocation20 languages
AR809
3
Showing 25 of 25 rows