<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="khelli-etal-2025-causes">
    <titleInfo>
        <title>What Causes Knowledge Loss in Multilingual Language Models?</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Maria</namePart>
        <namePart type="family">Khelli</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Samuel</namePart>
        <namePart type="family">Cahyawijaya</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Ayu</namePart>
        <namePart type="family">Purwarianti</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Genta</namePart>
        <namePart type="given">Indra</namePart>
        <namePart type="family">Winata</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2025-08</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the Fourth Workshop on NLP Applications to Field Linguistics</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Éric</namePart>
            <namePart type="family">Le Ferrand</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Elena</namePart>
            <namePart type="family">Klyachko</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Anna</namePart>
            <namePart type="family">Postnikova</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Tatiana</namePart>
            <namePart type="family">Shavrina</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Oleg</namePart>
            <namePart type="family">Serikov</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ekaterina</namePart>
            <namePart type="family">Voloshina</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ekaterina</namePart>
            <namePart type="family">Vylomova</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">979-8-89176-282-4</identifier>
    </relatedItem>
    <abstract>Cross-lingual transfer in natural language processing (NLP) models enhances multilingual performance by leveraging shared linguistic knowledge. However, traditional methods that process all data simultaneously often fail to mimic real-world scenarios, leading to challenges like catastrophic forgetting, where fine-tuning on new tasks degrades performance on previously learned ones. Our study explores this issue in multilingual contexts, focusing on linguistic differences affecting representational learning rather than just model parameters. We experiment with 52 languages using LoRA adapters of varying ranks to evaluate non-shared, partially shared, and fully shared parameters. Our aim is to see if parameter sharing through adapters can mitigate forgetting while preserving prior knowledge. We find that languages using non-Latin scripts are more susceptible to catastrophic forgetting, whereas those written in Latin script facilitate more effective cross-lingual transfer.</abstract>
    <identifier type="citekey">khelli-etal-2025-causes</identifier>
    <location>
        <url>https://aclanthology.org/2025.fieldmatters-1.2/</url>
    </location>
    <part>
        <date>2025-08</date>
        <extent unit="page">
            <start>15</start>
            <end>25</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T What Causes Knowledge Loss in Multilingual Language Models?
%A Khelli, Maria
%A Cahyawijaya, Samuel
%A Purwarianti, Ayu
%A Winata, Genta Indra
%Y Le Ferrand, Éric
%Y Klyachko, Elena
%Y Postnikova, Anna
%Y Shavrina, Tatiana
%Y Serikov, Oleg
%Y Voloshina, Ekaterina
%Y Vylomova, Ekaterina
%S Proceedings of the Fourth Workshop on NLP Applications to Field Linguistics
%D 2025
%8 August
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-282-4
%F khelli-etal-2025-causes
%X Cross-lingual transfer in natural language processing (NLP) models enhances multilingual performance by leveraging shared linguistic knowledge. However, traditional methods that process all data simultaneously often fail to mimic real-world scenarios, leading to challenges like catastrophic forgetting, where fine-tuning on new tasks degrades performance on previously learned ones. Our study explores this issue in multilingual contexts, focusing on linguistic differences affecting representational learning rather than just model parameters. We experiment with 52 languages using LoRA adapters of varying ranks to evaluate non-shared, partially shared, and fully shared parameters. Our aim is to see if parameter sharing through adapters can mitigate forgetting while preserving prior knowledge. We find that languages using non-Latin scripts are more susceptible to catastrophic forgetting, whereas those written in Latin script facilitate more effective cross-lingual transfer.
%U https://aclanthology.org/2025.fieldmatters-1.2/
%P 15-25

Markdown (Informal)

[What Causes Knowledge Loss in Multilingual Language Models?](https://aclanthology.org/2025.fieldmatters-1.2/) (Khelli et al., FieldMatters 2025)

• What Causes Knowledge Loss in Multilingual Language Models? (Khelli et al., FieldMatters 2025)

ACL

• Maria Khelli, Samuel Cahyawijaya, Ayu Purwarianti, and Genta Indra Winata. 2025. What Causes Knowledge Loss in Multilingual Language Models?. In Proceedings of the Fourth Workshop on NLP Applications to Field Linguistics, pages 15–25, Vienna, Austria. Association for Computational Linguistics.