arXivRyo Fukuda, Atsushi Ando, Hiroki Kanagawa, Takatomo Kano, Marc Delcroix, Naohiro Tawara, Yuya ChibaFri, Jul 31, 2026, 12:56 AM PDT
score 15.0
New Benchmark Tests AI That Listens and Speaks at Once
Original: M3-DuplexBench: A Multi-Turn, Multilingual, Multidomain Benchmark for Full-Duplex Spoken Dialogue Models
Source: arxiv.org ↗
Writing ELI5 summary…