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

Bilingual Full-Duplex-Bench

Benchmarks

Task NameDataset NameSOTA ResultTrend
User InterruptionBilingual Full-Duplex-Bench English
RL2.75
12
Overall EvaluationBilingual Full-Duplex-Bench English
Accuracy81.2
8
User BackchannelBilingual Full-Duplex-Bench English
RsR98
6
Turn TakingBilingual Full-Duplex-Bench English
TOR99.2
6
Pause HandlingBilingual Full-Duplex-Bench English
TOR98.3
6
User InterruptionBilingual Full-Duplex-Bench Chinese
RL1.63
4
Overall EvaluationBilingual Full-Duplex-Bench Chinese
Accuracy91.6
2
User BackchannelBilingual Full-Duplex-Bench Chinese
RsR80
2
Pause HandlingBilingual Full-Duplex-Bench Chinese
TOR4.2
2
Showing 9 of 9 rows