AI RESEARCH
RelayS2S: A Dual-Path Speculative Generation for Real-Time Dialogue
arXiv CS.AI
•
ArXi:2603.23346v1 Announce Type: new Real-time spoken dialogue systems face a fundamental tension between latency and response quality. End-to-end speech-to-speech (S2S) models respond immediately and naturally handle turn-taking, backchanneling, and interruption, but produce semantically weaker outputs. Cascaded pipelines (ASR -> LLM) deliver stronger responses at the cost of latency that grows with model size. We present RelayS2S, a hybrid architecture that runs two paths in parallel upon turn detection.