aiqbench analysis
This study investigates the emergent timing dynamics when full-duplex speech models converse autonomously, revealing that their turn-taking settles into a reactive loop with significant latency compared to human projection. The findings demonstrate that without human mediation to absorb errors, model-agent interactions exhibit coupled delays and empty run-up periods, highlighting a fundamental gap in simulating natural conversational flow.
Coupled but Late: Turn-Taking Between Full-Duplex Speech Models in Unscripted Dialogue
arXiv:2610.08683v1 Announce Type: new Abstract: Full-duplex speech models are trained to converse with a person, but they are increasingly made to converse with each other, in self-play data generation, agent societies, and model-based evaluation. In that loop no human absorbs a timing error: each model's turn-taking is the other's input. We ask what timing the loop settles into. Two PersonaPlex-7B instances exchange audio tokens on a shared clock in unscripted conversation, and one floor-transfer rule is applied to them and to Switchboard. Their timing is coupled: re-pairing speakers across conversations destroys it. But the floor changes hands late, at a median of 400-560 ms against 137 ms for humans, and the last 120 ms of the partner's turn, where human projection places a tenth of its transfers, holds 1% of theirs. Delaying one direction of the channel shifts the response one-for-one and leaves the run-up to it empty, consistent with a reactive wait after the perceived end rather than the turn-end projection human timing requires.
Abstract:Full-duplex speech models are trained to converse with a person, but they are increasingly made to converse with each other, in self-play data generation, agent societies, and model-based evaluation. In that loop no human absorbs a timing error: each model's turn-taking is the other's input. We ask what timing the loop settles into. Two PersonaPlex-7B instances exchange audio tokens on a shared clock in unscripted conversation, and one floor-transfer rule is applied to them and to Switchboard. Their timing is coupled: re-pairing speakers across conversations destroys it. But the floor changes hands late, at a median of 400-560 ms against 137 ms for humans, and the last 120 ms of the partner's turn, where human projection places a tenth of its transfers, holds 1% of theirs. Delaying one direction of the channel shifts the response one-for-one and leaves the run-up to it empty, consistent with a reactive wait after the perceived end rather than the turn-end projection human timing requires.
Artificial Intelligence (cs.AI)
arXiv:2610.08683 [cs.AI]
(or arXiv:2610.08683v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2610.08683 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 254 words
Related coverage
Reference-Guided Turn-Taking Style Transfer for Full-Duplex Spoken Language Models
arXiv · Oct 17, 2026
Links #5: 2026/07
LessWrong · Jul 31, 2026
Conversation Is a Two-Body Problem: Dyadic Evaluation of Full-Duplex Dialogue Models
arXiv · Oct 7, 2026
It’s about ethics in journalism, with Ben Smith
The Verge · Aug 17, 2026