@inproceedings{pham-etal-2026-language,
title = "Language Diffusion Models are Associative Memories Capable of Retrieving Unseen Data",
author = "Pham, Bao and
Zaki, Mohammed J and
Ambrogioni, Luca and
Krotov, Dmitry and
Negri, Matteo",
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.835/",
pages = "19377--19404",
ISBN = "979-8-89176-459-0",
abstract = "Understanding when language diffusion models memorize training data versus generalize remains an open challenge. We address it by framing Uniform-based Discrete Diffusion Models (UDDMs) as Associative Memories (AMs) where **both memorization and creativity coexist**. Unlike Hopfield networks, which require an energy function to produce attractors, we show that attractors can be formed via conditional likelihood maximization. We observe that UDDMs treat both training and unseen test sequences as fixed points for denoising while retaining the capacity to synthesize novel texts. We demonstrate that, as training set size increases, attractor basins around training samples shrink while those around unseen test samples expand until they become indistinguishable. This transition can be detected via the conditional entropy of tokens, which vanishes during memorization. Our framework provides both a theoretical foundation and a practical metric to elucidate the generative regimes of UDDMs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="pham-etal-2026-language">
<titleInfo>
<title>Language Diffusion Models are Associative Memories Capable of Retrieving Unseen Data</title>
</titleInfo>
<name type="personal">
<namePart type="given">Bao</namePart>
<namePart type="family">Pham</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammed</namePart>
<namePart type="given">J</namePart>
<namePart type="family">Zaki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Luca</namePart>
<namePart type="family">Ambrogioni</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dmitry</namePart>
<namePart type="family">Krotov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Matteo</namePart>
<namePart type="family">Negri</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>Understanding when language diffusion models memorize training data versus generalize remains an open challenge. We address it by framing Uniform-based Discrete Diffusion Models (UDDMs) as Associative Memories (AMs) where **both memorization and creativity coexist**. Unlike Hopfield networks, which require an energy function to produce attractors, we show that attractors can be formed via conditional likelihood maximization. We observe that UDDMs treat both training and unseen test sequences as fixed points for denoising while retaining the capacity to synthesize novel texts. We demonstrate that, as training set size increases, attractor basins around training samples shrink while those around unseen test samples expand until they become indistinguishable. This transition can be detected via the conditional entropy of tokens, which vanishes during memorization. Our framework provides both a theoretical foundation and a practical metric to elucidate the generative regimes of UDDMs.</abstract>
<identifier type="citekey">pham-etal-2026-language</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.835/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>19377</start>
<end>19404</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Language Diffusion Models are Associative Memories Capable of Retrieving Unseen Data
%A Pham, Bao
%A Zaki, Mohammed J.
%A Ambrogioni, Luca
%A Krotov, Dmitry
%A Negri, Matteo
%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 pham-etal-2026-language
%X Understanding when language diffusion models memorize training data versus generalize remains an open challenge. We address it by framing Uniform-based Discrete Diffusion Models (UDDMs) as Associative Memories (AMs) where **both memorization and creativity coexist**. Unlike Hopfield networks, which require an energy function to produce attractors, we show that attractors can be formed via conditional likelihood maximization. We observe that UDDMs treat both training and unseen test sequences as fixed points for denoising while retaining the capacity to synthesize novel texts. We demonstrate that, as training set size increases, attractor basins around training samples shrink while those around unseen test samples expand until they become indistinguishable. This transition can be detected via the conditional entropy of tokens, which vanishes during memorization. Our framework provides both a theoretical foundation and a practical metric to elucidate the generative regimes of UDDMs.
%U https://aclanthology.org/2026.emnlp-main.835/
%P 19377-19404
Markdown (Informal)
[Language Diffusion Models are Associative Memories Capable of Retrieving Unseen Data](https://aclanthology.org/2026.emnlp-main.835/) (Pham et al., EMNLP 2026)
ACL