HiThink Turn improves interruption timing in full-duplex conversations
HiThink Turn: An Intent-Aware Turn-Taking Control Module for Full-Duplex Dialogue
Human-Computer Interaction
Summary
Having conversations where both people can talk and listen at the same time is tricky because it's hard to know when to interrupt or wait. The authors created HiThink Turn, a system that watches the intentions behind what's being said and decides when to take turns without waiting for someone to finish completely. They trained it to recognize small chunks of speech that carry enough meaning to decide if an interruption is needed. This approach helps conversations flow faster and with fewer mistakes in timing interruptions.
What this means in practice
- •For voice assistant developers: Control turn-taking in voice assistants to allow timely and accurate user interruptions without waiting for full utterances to end.$Commercial implications: Enables improved real-time voice assistant interaction by allowing faster user interruption management.
- •For call center software engineers: Implement smarter conversation controls in call centers that detect when agents or customers intend to speak before current turns finish.
Authors
Feiyang Chen, Wenhan Yang, Bohan Wang, Xinjian Gao, Rongjunchen Zhang, Jun Wang, Xinhui Hu
Abstract
Full-duplex dialogue requires timely yet selective interruption handling, which end-of-turn prediction alone cannot achieve: complete utterances may need no response, while unfinished requests may warrant interruption. To address this challenge, we propose HiThink Turn, an intent-aware streaming turn-state predictor that separates response intent from semantic completeness and conditions decisions on system playback state. A key contribution is minimal intent-sufficient prefix supervision, constructed through LLM judgments and speech alignment, while training on audio truncated at chunk boundaries improves robustness to partial speech. These components support streaming inference with 240-ms audio chunks, enabling low-latency, accurate full-duplex turn control. Experiments show that HiThink Turn leads the compared methods in Easy Turn macro accuracy, Full-Duplex-Bench average interaction rate score (0.933), and non-target-speech average playback resume rate (0.735). Additionally, intent-prefix triggering raises interruption success from 89\% to 98\% and reduces mean stop latency by 60.9\%.