1 paper
Haixiao Gao, Yimin Zheng, Linyou Xiao +1
Full-duplex speech models support streaming interaction that listens and speaks at the same time. Serving them is governed by a strict, repeating deadline: conversation advances on…