1 paper
Chengqian Ma, Wenhao Feng, Weixuan Jin +7
Real-time full-duplex speech models can listen while speaking, enabling natural interaction without rigid turn boundaries. Existing benchmarks evaluate turn-taking, interruption ha…