@inproceedings{mehta-etal-2026-calm,
title = "{CALM}: Class-Conditional Sparse Attention Vectors for Large Audio-Language Models",
author = "Mehta, Videet and
Wang, Liming and
Kuehne, Hilde and
Feris, Rogerio and
Glass, James R. and
Mirza, Muhammad Jehanzeb",
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.10/",
pages = "234--245",
ISBN = "979-8-89176-459-0",
abstract = "Large audio-language models (LALMs) show strong zero-shot performance on tasks such as audio question answering, but still lag behind specialized systems on discriminative tasks such as audio classification. Recent work shows that sparse subsets of attention heads can serve as discriminative feature extractors for few-shot classification, but existing methods weight all selected heads equally. We propose $\textbf{C}$lass-Conditional Sparse $\textbf{A}$ttention Vectors for Large Audio-$\textbf{L}$anguage $\textbf{M}$odels, a few-shot classifier that learns class-conditional importance weights over attention heads. By treating heads as class-specific experts and aggregating them with reliability-weighted soft voting, CALM allows different classes to rely on different subsets of heads. Across diverse few-shot audio, audiovisual, visual, and spoofing benchmarks, CALM consistently outperforms state-of-the-art uniform-voting baselines, with gains of up to 14.52, 1.53, and 8.35 absolute points on audio, audiovisual, and spoofing tasks, respectively."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="mehta-etal-2026-calm">
<titleInfo>
<title>CALM: Class-Conditional Sparse Attention Vectors for Large Audio-Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Videet</namePart>
<namePart type="family">Mehta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Liming</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hilde</namePart>
<namePart type="family">Kuehne</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rogerio</namePart>
<namePart type="family">Feris</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">James</namePart>
<namePart type="given">R</namePart>
<namePart type="family">Glass</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Muhammad</namePart>
<namePart type="given">Jehanzeb</namePart>
<namePart type="family">Mirza</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>Large audio-language models (LALMs) show strong zero-shot performance on tasks such as audio question answering, but still lag behind specialized systems on discriminative tasks such as audio classification. Recent work shows that sparse subsets of attention heads can serve as discriminative feature extractors for few-shot classification, but existing methods weight all selected heads equally. We propose Class-Conditional Sparse Attention Vectors for Large Audio-Language Models, a few-shot classifier that learns class-conditional importance weights over attention heads. By treating heads as class-specific experts and aggregating them with reliability-weighted soft voting, CALM allows different classes to rely on different subsets of heads. Across diverse few-shot audio, audiovisual, visual, and spoofing benchmarks, CALM consistently outperforms state-of-the-art uniform-voting baselines, with gains of up to 14.52, 1.53, and 8.35 absolute points on audio, audiovisual, and spoofing tasks, respectively.</abstract>
<identifier type="citekey">mehta-etal-2026-calm</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.10/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>234</start>
<end>245</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CALM: Class-Conditional Sparse Attention Vectors for Large Audio-Language Models
%A Mehta, Videet
%A Wang, Liming
%A Kuehne, Hilde
%A Feris, Rogerio
%A Glass, James R.
%A Mirza, Muhammad Jehanzeb
%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 mehta-etal-2026-calm
%X Large audio-language models (LALMs) show strong zero-shot performance on tasks such as audio question answering, but still lag behind specialized systems on discriminative tasks such as audio classification. Recent work shows that sparse subsets of attention heads can serve as discriminative feature extractors for few-shot classification, but existing methods weight all selected heads equally. We propose Class-Conditional Sparse Attention Vectors for Large Audio-Language Models, a few-shot classifier that learns class-conditional importance weights over attention heads. By treating heads as class-specific experts and aggregating them with reliability-weighted soft voting, CALM allows different classes to rely on different subsets of heads. Across diverse few-shot audio, audiovisual, visual, and spoofing benchmarks, CALM consistently outperforms state-of-the-art uniform-voting baselines, with gains of up to 14.52, 1.53, and 8.35 absolute points on audio, audiovisual, and spoofing tasks, respectively.
%U https://aclanthology.org/2026.emnlp-main.10/
%P 234-245
Markdown (Informal)
[CALM: Class-Conditional Sparse Attention Vectors for Large Audio-Language Models](https://aclanthology.org/2026.emnlp-main.10/) (Mehta et al., EMNLP 2026)
ACL