<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yeh-etal-2026-learning">
    <titleInfo>
        <title>Learning Speech Representations with Variational Predictive Coding</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Sung-Lin</namePart>
        <namePart type="family">Yeh</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Peter</namePart>
        <namePart type="family">Bell</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Hao</namePart>
        <namePart type="family">Tang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <genre authority="bibutilsgt">journal article</genre>
    <relatedItem type="host">
        <titleInfo>
            <title>Transactions of the Association for Computational Linguistics</title>
        </titleInfo>
        <originInfo>
            <issuance>continuing</issuance>
            <publisher>MIT Press</publisher>
            <place>
                <placeTerm type="text">Cambridge, MA</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">periodical</genre>
        <genre authority="bibutilsgt">academic journal</genre>
    </relatedItem>
    <abstract>Despite being the best known objective for learning speech representations, the HuBERT objective has not been further developed and improved. We argue that it is the lack of an underlying principle that stalls the development, and, in this paper, we show that predictive coding under a variational view is the principle behind the HuBERT objective. Due to its generality, our formulation provides opportunities to improve parameterization and optimization, and we show two simple modifications that bring immediate improvements to the HuBERT objective. In addition, the predictive coding formulation has tight connections to various other objectives, such as APC, CPC, wav2vec, and BEST-RQ. Empirically, the improvement in pre-training brings significant improvements to four downstream tasks: phone classification, f0 tracking, speaker recognition, and automatic speech recognition, highlighting the importance of the predictive coding interpretation.</abstract>
    <identifier type="citekey">yeh-etal-2026-learning</identifier>
    <identifier type="doi">10.1162/tacl.a.738</identifier>
    <location>
        <url>https://aclanthology.org/2026.tacl-1.68/</url>
    </location>
    <part>
        <date>2026</date>
        <detail type="volume"><number>14</number></detail>
        <extent unit="page">
            <start>1510</start>
            <end>1525</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Journal Article
%T Learning Speech Representations with Variational Predictive Coding
%A Yeh, Sung-Lin
%A Bell, Peter
%A Tang, Hao
%J Transactions of the Association for Computational Linguistics
%D 2026
%V 14
%I MIT Press
%C Cambridge, MA
%F yeh-etal-2026-learning
%X Despite being the best known objective for learning speech representations, the HuBERT objective has not been further developed and improved. We argue that it is the lack of an underlying principle that stalls the development, and, in this paper, we show that predictive coding under a variational view is the principle behind the HuBERT objective. Due to its generality, our formulation provides opportunities to improve parameterization and optimization, and we show two simple modifications that bring immediate improvements to the HuBERT objective. In addition, the predictive coding formulation has tight connections to various other objectives, such as APC, CPC, wav2vec, and BEST-RQ. Empirically, the improvement in pre-training brings significant improvements to four downstream tasks: phone classification, f0 tracking, speaker recognition, and automatic speech recognition, highlighting the importance of the predictive coding interpretation.
%R 10.1162/tacl.a.738
%U https://aclanthology.org/2026.tacl-1.68/
%U https://doi.org/10.1162/tacl.a.738
%P 1510-1525

Markdown (Informal)

[Learning Speech Representations with Variational Predictive Coding](https://aclanthology.org/2026.tacl-1.68/) (Yeh et al., TACL 2026)

• Learning Speech Representations with Variational Predictive Coding (Yeh et al., TACL 2026)

ACL

• Sung-Lin Yeh, Peter Bell, and Hao Tang. 2026. Learning Speech Representations with Variational Predictive Coding. Transactions of the Association for Computational Linguistics, 14:1510–1525.