@inproceedings{zou-etal-2026-deciphering,
title = "Deciphering Cultural Representations in Large Language Models via Sparse Autoencoders",
author = "Zou, Chenye and
Jiao, Difan and
Hu, Lijie",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.278/",
pages = "5656--5677",
ISBN = "979-8-89176-395-1",
abstract = "Large Language Models (LLMs) are increasingly deployed worldwide, yet they exhibit strong Western-centric biases, and the internal mechanisms governing their cultural behaviors remain poorly understood. Prior work has identified so-called cultural neurons, but individual neurons are often polysemous, conflating abstract cultural knowledge with surface-level lexical cues due to superposition. We apply Sparse Autoencoders (SAEs) to decompose intermediate LLM activations into sparse, interpretable feature representations that disentangle these factors. This analysis reveals culturally selective features that remain invariant across paraphrasing and task formats, indicating abstraction beyond lexical correlations. Through targeted feature ablation, we provide causal evidence that these features are necessary for cultural reasoning: their removal selectively degrades performance on culturally conditioned tasks. Furthermore, we show that steering model activations along these feature directions is sufficient to systematically modulate cultural-related knowledge generation, without retraining. Together, our results offer the first causal evidence that LLMs encode cultural knowledge as decoupled semantic structures rather than surface patterns, enabling a scalable pathway toward cultural alignment through mechanistic intervention. Code is available at https://github.com/IAN-YE/Cultural-features-SAE."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zou-etal-2026-deciphering">
<titleInfo>
<title>Deciphering Cultural Representations in Large Language Models via Sparse Autoencoders</title>
</titleInfo>
<name type="personal">
<namePart type="given">Chenye</namePart>
<namePart type="family">Zou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Difan</namePart>
<namePart type="family">Jiao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lijie</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-395-1</identifier>
</relatedItem>
<abstract>Large Language Models (LLMs) are increasingly deployed worldwide, yet they exhibit strong Western-centric biases, and the internal mechanisms governing their cultural behaviors remain poorly understood. Prior work has identified so-called cultural neurons, but individual neurons are often polysemous, conflating abstract cultural knowledge with surface-level lexical cues due to superposition. We apply Sparse Autoencoders (SAEs) to decompose intermediate LLM activations into sparse, interpretable feature representations that disentangle these factors. This analysis reveals culturally selective features that remain invariant across paraphrasing and task formats, indicating abstraction beyond lexical correlations. Through targeted feature ablation, we provide causal evidence that these features are necessary for cultural reasoning: their removal selectively degrades performance on culturally conditioned tasks. Furthermore, we show that steering model activations along these feature directions is sufficient to systematically modulate cultural-related knowledge generation, without retraining. Together, our results offer the first causal evidence that LLMs encode cultural knowledge as decoupled semantic structures rather than surface patterns, enabling a scalable pathway toward cultural alignment through mechanistic intervention. Code is available at https://github.com/IAN-YE/Cultural-features-SAE.</abstract>
<identifier type="citekey">zou-etal-2026-deciphering</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.278/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>5656</start>
<end>5677</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Deciphering Cultural Representations in Large Language Models via Sparse Autoencoders
%A Zou, Chenye
%A Jiao, Difan
%A Hu, Lijie
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F zou-etal-2026-deciphering
%X Large Language Models (LLMs) are increasingly deployed worldwide, yet they exhibit strong Western-centric biases, and the internal mechanisms governing their cultural behaviors remain poorly understood. Prior work has identified so-called cultural neurons, but individual neurons are often polysemous, conflating abstract cultural knowledge with surface-level lexical cues due to superposition. We apply Sparse Autoencoders (SAEs) to decompose intermediate LLM activations into sparse, interpretable feature representations that disentangle these factors. This analysis reveals culturally selective features that remain invariant across paraphrasing and task formats, indicating abstraction beyond lexical correlations. Through targeted feature ablation, we provide causal evidence that these features are necessary for cultural reasoning: their removal selectively degrades performance on culturally conditioned tasks. Furthermore, we show that steering model activations along these feature directions is sufficient to systematically modulate cultural-related knowledge generation, without retraining. Together, our results offer the first causal evidence that LLMs encode cultural knowledge as decoupled semantic structures rather than surface patterns, enabling a scalable pathway toward cultural alignment through mechanistic intervention. Code is available at https://github.com/IAN-YE/Cultural-features-SAE.
%U https://aclanthology.org/2026.findings-acl.278/
%P 5656-5677
Markdown (Informal)
[Deciphering Cultural Representations in Large Language Models via Sparse Autoencoders](https://aclanthology.org/2026.findings-acl.278/) (Zou et al., Findings 2026)
ACL