<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ohashi-etal-2026-multi">
    <titleInfo>
        <title>Multi-Faceted Interactivity Alignment in Full-Duplex Speech Models</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Atsumoto</namePart>
        <namePart type="family">Ohashi</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Neil</namePart>
        <namePart type="family">Zeghidour</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Alexandre</namePart>
        <namePart type="family">Défossez</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Eugene</namePart>
        <namePart type="family">Kharitonov</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-10</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Hung-Yi</namePart>
            <namePart type="family">Lee</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Sunipa</namePart>
            <namePart type="family">Dev</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jessy</namePart>
            <namePart type="family">Li</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Desmond</namePart>
            <namePart type="family">Elliott</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Budapest, Hungary</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-459-0</identifier>
    </relatedItem>
    <abstract>Full-duplex spoken dialogue models can listen and speak simultaneously, making them a promising architecture for natural conversation. However, current models are trained solely with supervised learning through token-level likelihood maximization, which does not directly optimize interaction-level behaviors, causing interactivity issues such as excessive silence and ill-timed turn-taking. Recent work has applied reinforcement learning (RL) to improve interactivity, but existing methods address only a limited set of interactive behaviors in their rewards. In this work, we propose a post-training alignment method that comprehensively improves the interactivity of full-duplex spoken dialogue models through RL. We address the four canonical axes of interactivity: pause handling, turn-taking, backchanneling, and user interruption. For each axis, we extract short audio segments from human conversation corpora and optimize the model with axis-specific reward functions. An extra LLM-based reward for response quality prevents semantic degradation. We apply our method to two open-source models, Moshi and PersonaPlex, demonstrating consistent improvements in interactivity on both offline evaluation with pre-recorded audio and real-time multi-turn dialogue evaluation.</abstract>
    <identifier type="citekey">ohashi-etal-2026-multi</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.2331/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>52949</start>
            <end>52965</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Multi-Faceted Interactivity Alignment in Full-Duplex Speech Models
%A Ohashi, Atsumoto
%A Zeghidour, Neil
%A Défossez, Alexandre
%A Kharitonov, Eugene
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F ohashi-etal-2026-multi
%X Full-duplex spoken dialogue models can listen and speak simultaneously, making them a promising architecture for natural conversation. However, current models are trained solely with supervised learning through token-level likelihood maximization, which does not directly optimize interaction-level behaviors, causing interactivity issues such as excessive silence and ill-timed turn-taking. Recent work has applied reinforcement learning (RL) to improve interactivity, but existing methods address only a limited set of interactive behaviors in their rewards. In this work, we propose a post-training alignment method that comprehensively improves the interactivity of full-duplex spoken dialogue models through RL. We address the four canonical axes of interactivity: pause handling, turn-taking, backchanneling, and user interruption. For each axis, we extract short audio segments from human conversation corpora and optimize the model with axis-specific reward functions. An extra LLM-based reward for response quality prevents semantic degradation. We apply our method to two open-source models, Moshi and PersonaPlex, demonstrating consistent improvements in interactivity on both offline evaluation with pre-recorded audio and real-time multi-turn dialogue evaluation.
%U https://aclanthology.org/2026.emnlp-main.2331/
%P 52949-52965

Markdown (Informal)

[Multi-Faceted Interactivity Alignment in Full-Duplex Speech Models](https://aclanthology.org/2026.emnlp-main.2331/) (Ohashi et al., EMNLP 2026)

• Multi-Faceted Interactivity Alignment in Full-Duplex Speech Models (Ohashi et al., EMNLP 2026)

ACL

• Atsumoto Ohashi, Neil Zeghidour, Alexandre Défossez, and Eugene Kharitonov. 2026. Multi-Faceted Interactivity Alignment in Full-Duplex Speech Models. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 52949–52965, Budapest, Hungary. Association for Computational Linguistics.