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

BELLE

Benchmarks

Task NameDataset NameSOTA ResultTrend
Speech-to-speech conversationBELLE-eval-S2S (Zh→En)
C-MOS4.33
2
Speech-to-speech conversationBELLE-eval-S2S (Zh→Zh)
C-MOS4.42
2
Speech-to-speech conversationBELLE-eval-S2S (Zh) (test)
GPT Score (T)4.8
2
Chinese-to-English Speech-to-Speech ConversationBELLE eval S2S
GPT Score (T)3.53
2
Showing 4 of 4 rows