<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="mita-etal-2026-linguistically">
    <titleInfo>
        <title>Linguistically Inspired Multilingual Pre-Pretraining</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Masato</namePart>
        <namePart type="family">Mita</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Max</namePart>
        <namePart type="family">Müller-Eberstein</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Misaki</namePart>
        <namePart type="family">Yamaguchi</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yohei</namePart>
        <namePart type="family">Oseki</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-10</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the Second BabyLM Workshop</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Leshem</namePart>
            <namePart type="family">Choshen</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ryan</namePart>
            <namePart type="family">Cotterell</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Mustafa</namePart>
            <namePart type="given">Omer</namePart>
            <namePart type="family">Gul</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jaap</namePart>
            <namePart type="family">Jumelet</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Tal</namePart>
            <namePart type="family">Linzen</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Aaron</namePart>
            <namePart type="family">Mueller</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Suchir</namePart>
            <namePart type="family">Salhan</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Raj</namePart>
            <namePart type="given">Sanjay</namePart>
            <namePart type="family">Shah</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Alex</namePart>
            <namePart type="family">Warstadt</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ethan</namePart>
            <namePart type="given">Gotlieb</namePart>
            <namePart type="family">Wilcox</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Budapest, Hungary</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-475-0</identifier>
    </relatedItem>
    <abstract>Encoding Universal Grammar principles (UG) during pre-pretraining (PPT) has been shown to produce inductive biases in language models, which facilitate natural language acquisition. While UG principles should be language-agnostic, each language follows different typological parameters. To empirically validate this theory, we operationalize three levels of typological flexibility via PPT: non-typological k-shuffle Dyck, as well as MP-STRUCT, which encodes the Minimalist Program operations MERGE, AGREE, MOVE following fixed typological parameters. Additionally, we introduce MP-STRUCT BCC (BCC), which parameterizes typological features following the Borer–Chomsky Conjecture to allow for a more typologically flexible parametrization. Through a multilingual, subtask-level evaluation across 207 grammatical tasks for English, Dutch, and Chinese, we find that principles-driven PPT significantly improves average accuracy, but selectively degrades language-specific phenomena such as wh-island constructions. In contrast, structure-only PPT, as well as typologically flexible PPT eliminate wh-island degradation while preserving overall gains, demonstrating that typological parameterization offers a path toward multilingual syntactic priors. Exploratory model-internal analyses are consistent with these findings at the levels of output probabilities, attention structure, as well as cross-lingual representation alignment.</abstract>
    <identifier type="citekey">mita-etal-2026-linguistically</identifier>
    <location>
        <url>https://aclanthology.org/2026.babylm-main.15/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>187</start>
            <end>198</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Linguistically Inspired Multilingual Pre-Pretraining
%A Mita, Masato
%A Müller-Eberstein, Max
%A Yamaguchi, Misaki
%A Oseki, Yohei
%Y Choshen, Leshem
%Y Cotterell, Ryan
%Y Gul, Mustafa Omer
%Y Jumelet, Jaap
%Y Linzen, Tal
%Y Mueller, Aaron
%Y Salhan, Suchir
%Y Shah, Raj Sanjay
%Y Warstadt, Alex
%Y Wilcox, Ethan Gotlieb
%S Proceedings of the Second BabyLM Workshop
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-475-0
%F mita-etal-2026-linguistically
%X Encoding Universal Grammar principles (UG) during pre-pretraining (PPT) has been shown to produce inductive biases in language models, which facilitate natural language acquisition. While UG principles should be language-agnostic, each language follows different typological parameters. To empirically validate this theory, we operationalize three levels of typological flexibility via PPT: non-typological k-shuffle Dyck, as well as MP-STRUCT, which encodes the Minimalist Program operations MERGE, AGREE, MOVE following fixed typological parameters. Additionally, we introduce MP-STRUCT BCC (BCC), which parameterizes typological features following the Borer–Chomsky Conjecture to allow for a more typologically flexible parametrization. Through a multilingual, subtask-level evaluation across 207 grammatical tasks for English, Dutch, and Chinese, we find that principles-driven PPT significantly improves average accuracy, but selectively degrades language-specific phenomena such as wh-island constructions. In contrast, structure-only PPT, as well as typologically flexible PPT eliminate wh-island degradation while preserving overall gains, demonstrating that typological parameterization offers a path toward multilingual syntactic priors. Exploratory model-internal analyses are consistent with these findings at the levels of output probabilities, attention structure, as well as cross-lingual representation alignment.
%U https://aclanthology.org/2026.babylm-main.15/
%P 187-198

Markdown (Informal)

[Linguistically Inspired Multilingual Pre-Pretraining](https://aclanthology.org/2026.babylm-main.15/) (Mita et al., BabyLM 2026)

• Linguistically Inspired Multilingual Pre-Pretraining (Mita et al., BabyLM 2026)

ACL

• Masato Mita, Max Müller-Eberstein, Misaki Yamaguchi, and Yohei Oseki. 2026. Linguistically Inspired Multilingual Pre-Pretraining. In Proceedings of the Second BabyLM Workshop, pages 187–198, Budapest, Hungary. Association for Computational Linguistics.