<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="shi-etal-2026-additive">
    <titleInfo>
        <title>On the Additive Compositionality of Task Vectors in Vision–Language Models</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Yuting</namePart>
        <namePart type="family">Shi</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Houjing</namePart>
        <namePart type="family">Wei</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Naoya</namePart>
        <namePart type="family">Inoue</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-03</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Vera</namePart>
            <namePart type="family">Demberg</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Kentaro</namePart>
            <namePart type="family">Inui</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Lluís</namePart>
            <namePart type="family">Marquez</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Rabat, Morocco</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-381-4</identifier>
    </relatedItem>
    <abstract>In-context learning (ICL) in large language models (LLMs) has been shown to operate through task vectors—the representation that summarizes the mapping induced by in-context demonstrations and can be composed by simple arithmetic operations. While this phenomenon is well studied in LLMs, its extension to vision-language models (VLMs) remains underexplored. In this work, we systematically examine the additive compositionality of in-context task vectors in VLMs, extracted from text-side hidden representations. Specifically, we construct compositional visual reasoning tasks with clearly defined subtasks and extract task vectors from few-shot demonstrations. Empirical experiments show that the vector for a complex task can be approximated by adding the vectors of its constituent subtasks. Beyond this, we analyze token-level contextual embeddings and show that additive composition arises because complex-task representations emerge as the superposition of atomic subtask components, preserving semantic structure within the model’s activation space.</abstract>
    <identifier type="citekey">shi-etal-2026-additive</identifier>
    <location>
        <url>https://aclanthology.org/2026.eacl-short.38/</url>
    </location>
    <part>
        <date>2026-03</date>
        <extent unit="page">
            <start>513</start>
            <end>521</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T On the Additive Compositionality of Task Vectors in Vision–Language Models
%A Shi, Yuting
%A Wei, Houjing
%A Inoue, Naoya
%Y Demberg, Vera
%Y Inui, Kentaro
%Y Marquez, Lluís
%S Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)
%D 2026
%8 March
%I Association for Computational Linguistics
%C Rabat, Morocco
%@ 979-8-89176-381-4
%F shi-etal-2026-additive
%X In-context learning (ICL) in large language models (LLMs) has been shown to operate through task vectors—the representation that summarizes the mapping induced by in-context demonstrations and can be composed by simple arithmetic operations. While this phenomenon is well studied in LLMs, its extension to vision-language models (VLMs) remains underexplored. In this work, we systematically examine the additive compositionality of in-context task vectors in VLMs, extracted from text-side hidden representations. Specifically, we construct compositional visual reasoning tasks with clearly defined subtasks and extract task vectors from few-shot demonstrations. Empirical experiments show that the vector for a complex task can be approximated by adding the vectors of its constituent subtasks. Beyond this, we analyze token-level contextual embeddings and show that additive composition arises because complex-task representations emerge as the superposition of atomic subtask components, preserving semantic structure within the model’s activation space.
%U https://aclanthology.org/2026.eacl-short.38/
%P 513-521

Markdown (Informal)

[On the Additive Compositionality of Task Vectors in Vision–Language Models](https://aclanthology.org/2026.eacl-short.38/) (Shi et al., EACL 2026)

• On the Additive Compositionality of Task Vectors in Vision–Language Models (Shi et al., EACL 2026)

ACL

• Yuting Shi, Houjing Wei, and Naoya Inoue. 2026. On the Additive Compositionality of Task Vectors in Vision–Language Models. In Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers), pages 513–521, Rabat, Morocco. Association for Computational Linguistics.