Title: From Turn-Taking to Synchronous Dialogue: A Survey of Full-Duplex Spoken Language Models Show Chinese title

Title: 从轮流对话到同步对话：全双工口语语言模型的综述

Abstract: True Full-Duplex (TFD) voice communication--enabling simultaneous listening and speaking with natural turn-taking, overlapping speech, and interruptions--represents a critical milestone toward human-like AI interaction. This survey comprehensively reviews Full-Duplex Spoken Language Models (FD-SLMs) in the LLM era. We establish a taxonomy distinguishing Engineered Synchronization (modular architectures) from Learned Synchronization (end-to-end architectures), and unify fragmented evaluation approaches into a framework encompassing Temporal Dynamics, Behavioral Arbitration, Semantic Coherence, and Acoustic Performance. Through comparative analysis of mainstream FD-SLMs, we identify fundamental challenges: synchronous data scarcity, architectural divergence, and evaluation gaps, providing a roadmap for advancing human-AI communication. Show Chinese abstract

Abstract: 真全双工（TFD）语音通信——实现同时聆听和说话，具有自然的轮流交谈、重叠语音和中断——标志着向类人AI交互的一个关键里程碑。 本综述全面回顾了大语言模型时代中的全双工口语语言模型（FD-SLMs）。 我们建立了一个分类体系，区分工程同步（模块化架构）与学习同步（端到端架构），并将碎片化的评估方法统一到一个涵盖时间动态、行为仲裁、语义连贯性和声学性能的框架中。 通过主流FD-SLMs的比较分析，我们识别出基本挑战：同步数据稀缺、架构分歧和评估差距，为推进人机通信提供了路线图。