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

MCIF

Benchmarks

Task NameDataset NameSOTA ResultTrend
Speech TranslationMCIF long track
COMET Score83.75
20
Automatic Speech RecognitionMCIF long track
WER5.88
20
Spoken SummarizationMCIF long track
BERTScore29.74
20
Spoken Question AnsweringMCIF long track
BERTScore40.86
20
Speech Question AnsweringMCIF short-form
BERTScore (SQA)45.13
10
Machine TranslationMCIF en-it
BLEU38.14
9
Machine TranslationMCIF en-de
BLEU30.88
9
Offline Speech TranslationMCIF (test)
XCOMET-XL86.59
9
Speech TranslationMCIF short-form
ST COMET78.69
8
Simultaneous Speech TranslationMCIF En-It (dev)
BLEU43.56
7
Simultaneous Speech TranslationMCIF En-De (dev)
BLEU31.73
7
Speech Question AnsweringMCIF constrained short audio
BERTScore-F1 (EN)34.6
4
Automatic Speech RecognitionMCIF constrained, short audio
WER21.49
4
Automatic Speech RecognitionMCIF short-form
ASR Accuracy88.77
4
Speech Instruction FollowingMCIF short-form
SIFS Score2.0708
2
Showing 15 of 15 rows