@inproceedings{yang-chang-2026-gated,
title = "Gated Contrastive Decoding via Local Candidate Geometry for Audio-Visual Speech Recognition",
author = "Yang, Da-Hee and
Chang, Joon-Hyuk",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1035/",
pages = "23767--23778",
ISBN = "979-8-89176-459-0",
abstract = "Contrastive decoding (CD) provides an effective inference-time strategy for audio-visual speech recognition (AVSR) by contrasting audio-visual and audio-only predictive distributions without retraining the underlying model. Existing approaches apply a fixed contrastive term at every decoding step, despite substantial variation in how informative the contrastive signal is across local decoding contexts. In AVSR, a fixed adjustment that benefits some decoding steps may provide little value or introduce instability in others, making uniform CD sensitive to the contrastive coefficient $\alpha$ and limiting its reliability across varying acoustic conditions. We propose Gated CD (GCD), a training-free framework that adapts the contrastive term according to local candidate geometry at each decoding step. Since contrastive adjustment primarily affects ranking among nearby competing tokens, GCD estimates local reliability through two complementary quantities {---} a margin support term and a dispersion risk term {---} that together capture whether the contrastive signal is directionally consistent and locally stable. These are combined into a continuous gate that adaptively modulates the contrastive term during decoding. Experiments on LRS3 and LRS2 across multiple noise conditions and model scales demonstrate that GCD consistently improves over both standard AVSR decoding and uniform CD, while substantially reducing sensitivity to $\alpha$, improving robustness across acoustic conditions."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yang-chang-2026-gated">
<titleInfo>
<title>Gated Contrastive Decoding via Local Candidate Geometry for Audio-Visual Speech Recognition</title>
</titleInfo>
<name type="personal">
<namePart type="given">Da-Hee</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joon-Hyuk</namePart>
<namePart type="family">Chang</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>Contrastive decoding (CD) provides an effective inference-time strategy for audio-visual speech recognition (AVSR) by contrasting audio-visual and audio-only predictive distributions without retraining the underlying model. Existing approaches apply a fixed contrastive term at every decoding step, despite substantial variation in how informative the contrastive signal is across local decoding contexts. In AVSR, a fixed adjustment that benefits some decoding steps may provide little value or introduce instability in others, making uniform CD sensitive to the contrastive coefficient α and limiting its reliability across varying acoustic conditions. We propose Gated CD (GCD), a training-free framework that adapts the contrastive term according to local candidate geometry at each decoding step. Since contrastive adjustment primarily affects ranking among nearby competing tokens, GCD estimates local reliability through two complementary quantities — a margin support term and a dispersion risk term — that together capture whether the contrastive signal is directionally consistent and locally stable. These are combined into a continuous gate that adaptively modulates the contrastive term during decoding. Experiments on LRS3 and LRS2 across multiple noise conditions and model scales demonstrate that GCD consistently improves over both standard AVSR decoding and uniform CD, while substantially reducing sensitivity to α, improving robustness across acoustic conditions.</abstract>
<identifier type="citekey">yang-chang-2026-gated</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1035/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>23767</start>
<end>23778</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Gated Contrastive Decoding via Local Candidate Geometry for Audio-Visual Speech Recognition
%A Yang, Da-Hee
%A Chang, Joon-Hyuk
%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 yang-chang-2026-gated
%X Contrastive decoding (CD) provides an effective inference-time strategy for audio-visual speech recognition (AVSR) by contrasting audio-visual and audio-only predictive distributions without retraining the underlying model. Existing approaches apply a fixed contrastive term at every decoding step, despite substantial variation in how informative the contrastive signal is across local decoding contexts. In AVSR, a fixed adjustment that benefits some decoding steps may provide little value or introduce instability in others, making uniform CD sensitive to the contrastive coefficient α and limiting its reliability across varying acoustic conditions. We propose Gated CD (GCD), a training-free framework that adapts the contrastive term according to local candidate geometry at each decoding step. Since contrastive adjustment primarily affects ranking among nearby competing tokens, GCD estimates local reliability through two complementary quantities — a margin support term and a dispersion risk term — that together capture whether the contrastive signal is directionally consistent and locally stable. These are combined into a continuous gate that adaptively modulates the contrastive term during decoding. Experiments on LRS3 and LRS2 across multiple noise conditions and model scales demonstrate that GCD consistently improves over both standard AVSR decoding and uniform CD, while substantially reducing sensitivity to α, improving robustness across acoustic conditions.
%U https://aclanthology.org/2026.emnlp-main.1035/
%P 23767-23778
Markdown (Informal)
[Gated Contrastive Decoding via Local Candidate Geometry for Audio-Visual Speech Recognition](https://aclanthology.org/2026.emnlp-main.1035/) (Yang & Chang, EMNLP 2026)
ACL