<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="dumas-etal-2025-separating">
    <titleInfo>
        <title>Separating Tongue from Thought: Activation Patching Reveals Language-Agnostic Concept Representations in Transformers</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Clément</namePart>
        <namePart type="family">Dumas</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Chris</namePart>
        <namePart type="family">Wendler</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Veniamin</namePart>
        <namePart type="family">Veselovsky</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Giovanni</namePart>
        <namePart type="family">Monea</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Robert</namePart>
        <namePart type="family">West</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2025-07</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Wanxiang</namePart>
            <namePart type="family">Che</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Joyce</namePart>
            <namePart type="family">Nabende</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ekaterina</namePart>
            <namePart type="family">Shutova</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Mohammad</namePart>
            <namePart type="given">Taher</namePart>
            <namePart type="family">Pilehvar</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Vienna, Austria</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-251-0</identifier>
    </relatedItem>
    <abstract>A central question in multilingual language modeling is whether large language models (LLMs) develop a universal concept representation, disentangled from specific languages. In this paper, we address this question by analyzing latent representations (latents) during a word-translation task in transformer-based LLMs. We strategically extract latents from a source translation prompt and insert them into the forward pass on a target translation prompt. By doing so, we find that the output language is encoded in the latent at an earlier layer than the concept to be translated. Building on this insight, we conduct two key experiments. First, we demonstrate that we can change the concept without changing the language and vice versa through activation patching alone. Second, we show that patching with the mean representation of a concept across different languages does not affect the models’ ability to translate it, but instead improves it. Finally, we generalize to multi-token generation and demonstrate that the model can generate natural language description of those mean representations. Our results provide evidence for the existence of language-agnostic concept representations within the investigated models.</abstract>
    <identifier type="citekey">dumas-etal-2025-separating</identifier>
    <identifier type="doi">10.18653/v1/2025.acl-long.1536</identifier>
    <location>
        <url>https://aclanthology.org/2025.acl-long.1536/</url>
    </location>
    <part>
        <date>2025-07</date>
        <extent unit="page">
            <start>31822</start>
            <end>31841</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Separating Tongue from Thought: Activation Patching Reveals Language-Agnostic Concept Representations in Transformers
%A Dumas, Clément
%A Wendler, Chris
%A Veselovsky, Veniamin
%A Monea, Giovanni
%A West, Robert
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-251-0
%F dumas-etal-2025-separating
%X A central question in multilingual language modeling is whether large language models (LLMs) develop a universal concept representation, disentangled from specific languages. In this paper, we address this question by analyzing latent representations (latents) during a word-translation task in transformer-based LLMs. We strategically extract latents from a source translation prompt and insert them into the forward pass on a target translation prompt. By doing so, we find that the output language is encoded in the latent at an earlier layer than the concept to be translated. Building on this insight, we conduct two key experiments. First, we demonstrate that we can change the concept without changing the language and vice versa through activation patching alone. Second, we show that patching with the mean representation of a concept across different languages does not affect the models’ ability to translate it, but instead improves it. Finally, we generalize to multi-token generation and demonstrate that the model can generate natural language description of those mean representations. Our results provide evidence for the existence of language-agnostic concept representations within the investigated models.
%R 10.18653/v1/2025.acl-long.1536
%U https://aclanthology.org/2025.acl-long.1536/
%U https://doi.org/10.18653/v1/2025.acl-long.1536
%P 31822-31841

Markdown (Informal)

[Separating Tongue from Thought: Activation Patching Reveals Language-Agnostic Concept Representations in Transformers](https://aclanthology.org/2025.acl-long.1536/) (Dumas et al., ACL 2025)

• Separating Tongue from Thought: Activation Patching Reveals Language-Agnostic Concept Representations in Transformers (Dumas et al., ACL 2025)

ACL

• Clément Dumas, Chris Wendler, Veniamin Veselovsky, Giovanni Monea, and Robert West. 2025. Separating Tongue from Thought: Activation Patching Reveals Language-Agnostic Concept Representations in Transformers. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 31822–31841, Vienna, Austria. Association for Computational Linguistics.