On the Limits of Model Merging for Multilinguality in Pre-Training

Seth Aycock; Fedor Vitiugin; Aleksandr Umnov; Christof Monz; Khalil Sima’an

On the Limits of Model Merging for Multilinguality in Pre-Training

Seth Aycock, Fedor Vitiugin, Aleksandr Umnov, Christof Monz, Khalil Sima’an

Correct Metadata for

Use this form to create a GitHub issue with structured data describing the correction. You will need a GitHub account. Once you create that issue, the correction will be reviewed by a staff member.

⚠️ Mobile Users: Submitting this form to create a new issue will only work with github.com, not the GitHub Mobile app.

Important: The Anthology treat PDFs as authoritative. Please use this form only to correct data that is out of line with the PDF. See our corrections guidelines if you need to change the PDF.

Title Adjust the title. Retain tags such as <fixed-case>.

Authors Adjust author names and order to match the PDF.

Abstract Correct abstract if needed. Retain XML formatting tags such as <tex-math>. You may use <b>...</b> for bold, <i>...</i> for italic, and <url>...</url> for URLs.

Verification against PDF Ensure that the new title/authors match the snapshot below. (If there is no snapshot or it is too small, consult the PDF.)

Authors concatenated from the text boxes above:

ALL author names match the snapshot above—including middle initials, hyphens, and accents.

Abstract

Endowing models with consistent multilingual performance can be achieved by _mixing_ pre-training data, or post-training approaches such as language-specific model _merging_. In this work, we test whether merging can be applied to monolingually pre-trained models. We conduct a controlled study on the efficacy of mixed, merged, and monolingual pre-training setups. We find that while monolingual pre-training results in strong in-language performance, merging any combination of monolingual models leads to performance collapse due to interference. Our analysis suggests representational similarity is a prerequisite for model merging. We therefore conclude that the flexibility of merging in fine-tuning does not extend trivially to language-specific pre-training.

Anthology ID:: 2026.mellm-1.15
Volume:: Proceedings of the 1st Workshop on Multilinguality in the Era of Large Language Models (MeLLM 2026)
Month:: July
Year:: 2026
Address:: San Diego, United States
Editors:: Kaiyu Huang, Fengran Mo, Pinzhen Chen, Meng Jiang
Venues:: MeLLM | WS
SIG:
Publisher:: Association for Computational Linguistics
Note:
Pages:: 159–169
Language:
URL:: https://aclanthology.org/2026.mellm-1.15/
DOI:
Bibkey:
Cite (ACL):: Seth Aycock, Fedor Vitiugin, Aleksandr Umnov, Christof Monz, and Khalil Sima’an. 2026. On the Limits of Model Merging for Multilinguality in Pre-Training. In Proceedings of the 1st Workshop on Multilinguality in the Era of Large Language Models (MeLLM 2026), pages 159–169, San Diego, United States. Association for Computational Linguistics.
Cite (Informal):: On the Limits of Model Merging for Multilinguality in Pre-Training (Aycock et al., MeLLM 2026)
Copy Citation:
PDF:: https://aclanthology.org/2026.mellm-1.15.pdf

PDF Cite Search Fix data

Export citation

BibTeX
MODS XML
Endnote
Preformatted

@inproceedings{aycock-etal-2026-limits,
    title = "On the Limits of Model Merging for Multilinguality in Pre-Training",
    author = "Aycock, Seth  and
      Vitiugin, Fedor  and
      Umnov, Aleksandr  and
      Monz, Christof  and
      Sima{'}an, Khalil",
    editor = "Huang, Kaiyu  and
      Mo, Fengran  and
      Chen, Pinzhen  and
      Jiang, Meng",
    booktitle = "Proceedings of the 1st Workshop on Multilinguality in the Era of Large Language Models ({M}e{LLM} 2026)",
    month = jul,
    year = "2026",
    address = "San Diego, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2026.mellm-1.15/",
    pages = "159--169",
    ISBN = "979-8-89176-430-9",
    abstract = "Endowing models with consistent multilingual performance can be achieved by {\_}mixing{\_} pre-training data, or post-training approaches such as language-specific model {\_}merging{\_}. In this work, we test whether merging can be applied to monolingually pre-trained models. We conduct a controlled study on the efficacy of mixed, merged, and monolingual pre-training setups. We find that while monolingual pre-training results in strong in-language performance, merging any combination of monolingual models leads to performance collapse due to interference. Our analysis suggests representational similarity is a prerequisite for model merging. We therefore conclude that the flexibility of merging in fine-tuning does not extend trivially to language-specific pre-training."
}

Download as File

<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="aycock-etal-2026-limits">
    <titleInfo>
        <title>On the Limits of Model Merging for Multilinguality in Pre-Training</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Seth</namePart>
        <namePart type="family">Aycock</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Fedor</namePart>
        <namePart type="family">Vitiugin</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Aleksandr</namePart>
        <namePart type="family">Umnov</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Christof</namePart>
        <namePart type="family">Monz</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Khalil</namePart>
        <namePart type="family">Sima’an</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>Proceedings of the 1st Workshop on Multilinguality in the Era of Large Language Models (MeLLM 2026)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Kaiyu</namePart>
            <namePart type="family">Huang</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Fengran</namePart>
            <namePart type="family">Mo</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Pinzhen</namePart>
            <namePart type="family">Chen</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Meng</namePart>
            <namePart type="family">Jiang</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">San Diego, United States</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-430-9</identifier>
    </relatedItem>
    <abstract>Endowing models with consistent multilingual performance can be achieved by _mixing_ pre-training data, or post-training approaches such as language-specific model _merging_. In this work, we test whether merging can be applied to monolingually pre-trained models. We conduct a controlled study on the efficacy of mixed, merged, and monolingual pre-training setups. We find that while monolingual pre-training results in strong in-language performance, merging any combination of monolingual models leads to performance collapse due to interference. Our analysis suggests representational similarity is a prerequisite for model merging. We therefore conclude that the flexibility of merging in fine-tuning does not extend trivially to language-specific pre-training.</abstract>
    <identifier type="citekey">aycock-etal-2026-limits</identifier>
    <location>
        <url>https://aclanthology.org/2026.mellm-1.15/</url>
    </location>
    <part>
        <date>2026-07</date>
        <extent unit="page">
            <start>159</start>
            <end>169</end>
        </extent>
    </part>
</mods>
</modsCollection>

Download as File

%0 Conference Proceedings
%T On the Limits of Model Merging for Multilinguality in Pre-Training
%A Aycock, Seth
%A Vitiugin, Fedor
%A Umnov, Aleksandr
%A Monz, Christof
%A Sima’an, Khalil
%Y Huang, Kaiyu
%Y Mo, Fengran
%Y Chen, Pinzhen
%Y Jiang, Meng
%S Proceedings of the 1st Workshop on Multilinguality in the Era of Large Language Models (MeLLM 2026)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, United States
%@ 979-8-89176-430-9
%F aycock-etal-2026-limits
%X Endowing models with consistent multilingual performance can be achieved by _mixing_ pre-training data, or post-training approaches such as language-specific model _merging_. In this work, we test whether merging can be applied to monolingually pre-trained models. We conduct a controlled study on the efficacy of mixed, merged, and monolingual pre-training setups. We find that while monolingual pre-training results in strong in-language performance, merging any combination of monolingual models leads to performance collapse due to interference. Our analysis suggests representational similarity is a prerequisite for model merging. We therefore conclude that the flexibility of merging in fine-tuning does not extend trivially to language-specific pre-training.
%U https://aclanthology.org/2026.mellm-1.15/
%P 159-169

Download as File

Markdown (Informal)

[On the Limits of Model Merging for Multilinguality in Pre-Training](https://aclanthology.org/2026.mellm-1.15/) (Aycock et al., MeLLM 2026)

On the Limits of Model Merging for Multilinguality in Pre-Training (Aycock et al., MeLLM 2026)

ACL

Seth Aycock, Fedor Vitiugin, Aleksandr Umnov, Christof Monz, and Khalil Sima’an. 2026. On the Limits of Model Merging for Multilinguality in Pre-Training. In Proceedings of the 1st Workshop on Multilinguality in the Era of Large Language Models (MeLLM 2026), pages 159–169, San Diego, United States. Association for Computational Linguistics.