Full-Duplex-Bench
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Full-Duplex-Bench Candor
1TOR
19
Full-Duplex-Bench
4.59GPT-4o Score
18
Full-Duplex-Bench
99.2TOR
17
Full-Duplex-Bench
0.01Synthetic TOR for Pause Handling
15
Full-Duplex Bench v1.0
0.983TOR
11
Full-Duplex-Bench
100TOR
11
Full-Duplex-Bench Synthetic
99TOR
11
Full-Duplex-Bench v1
93.4TOR (Synthetic Pause)
9
Full-Duplex-Bench v1.5
0.68Stop Latency (Mean)
8
Full-Duplex-Bench 1.0
1TOR
8
Full-Duplex-Bench Background Speech 1.5
93Respond Rate
7
Full-Duplex-Bench 1.5 (Talking to Other)
91Response Rate
7
Full-Duplex-Bench User Backchannel 1.5
7Respond Rate
7
Full-Duplex-Bench User Interruption 1.5
78Response Rate
7
Full-Duplex-Bench 1.0
1TOR
7
Full-Duplex-Bench
57SSIM
5
Full-Duplex-Bench
3.9DMOS
5
Full-Duplex Bench v1.5
7Respond Rate
4
Full-Duplex Bench v1.5
72Respond Rate
4
Full-Duplex-Bench User Interruption v1.5
0.97STOI
2
Full-Duplex-Bench User Backchannel v1.5
91STOI
2
Full-Duplex-Bench Talking to Other v1.5
0.96STOI
2
Full-Duplex-Bench Background Speech v1.5
0.98STOI
2
Full-Duplex-Bench Bilingual Chinese
99.4TOR
2
Full-Duplex-Bench EN
205Latency (ms)
1