<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="orten-etal-2025-neuron">
    <titleInfo>
        <title>Neuron-Level Language Tag Injection Improves Zero-Shot Translation Performance</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Jay</namePart>
        <namePart type="family">Orten</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Ammon</namePart>
        <namePart type="family">Shurtz</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Nancy</namePart>
        <namePart type="family">Fulda</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Stephen</namePart>
        <namePart type="given">D</namePart>
        <namePart type="family">Richardson</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 4: Student Research Workshop)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Jin</namePart>
            <namePart type="family">Zhao</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Mingyang</namePart>
            <namePart type="family">Wang</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Zhu</namePart>
            <namePart type="family">Liu</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">979-8-89176-254-1</identifier>
    </relatedItem>
    <abstract>Language tagging, a method whereby source and target inputs are prefixed with a unique language token, has become the de facto standard for conditioning Multilingual Neural Machine Translation (MNMT) models on specific language directions. This conditioning can manifest effective zero-shot translation abilities in MT models at scale for many languages. Expanding on previous work, we propose a novel method of language tagging for MNMT, injection, in which the embedded representation of a language token is concatenated to the input of every linear layer. We explore a variety of different tagging methods, with and without injection, showing that injection improves zero-shot translation performance with up to a 2+ BLEU score point gain for certain language directions in our dataset.</abstract>
    <identifier type="citekey">orten-etal-2025-neuron</identifier>
    <identifier type="doi">10.18653/v1/2025.acl-srw.13</identifier>
    <location>
        <url>https://aclanthology.org/2025.acl-srw.13/</url>
    </location>
    <part>
        <date>2025-07</date>
        <extent unit="page">
            <start>203</start>
            <end>212</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Neuron-Level Language Tag Injection Improves Zero-Shot Translation Performance
%A Orten, Jay
%A Shurtz, Ammon
%A Fulda, Nancy
%A Richardson, Stephen D.
%Y Zhao, Jin
%Y Wang, Mingyang
%Y Liu, Zhu
%S Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 4: Student Research Workshop)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-254-1
%F orten-etal-2025-neuron
%X Language tagging, a method whereby source and target inputs are prefixed with a unique language token, has become the de facto standard for conditioning Multilingual Neural Machine Translation (MNMT) models on specific language directions. This conditioning can manifest effective zero-shot translation abilities in MT models at scale for many languages. Expanding on previous work, we propose a novel method of language tagging for MNMT, injection, in which the embedded representation of a language token is concatenated to the input of every linear layer. We explore a variety of different tagging methods, with and without injection, showing that injection improves zero-shot translation performance with up to a 2+ BLEU score point gain for certain language directions in our dataset.
%R 10.18653/v1/2025.acl-srw.13
%U https://aclanthology.org/2025.acl-srw.13/
%U https://doi.org/10.18653/v1/2025.acl-srw.13
%P 203-212

Markdown (Informal)

[Neuron-Level Language Tag Injection Improves Zero-Shot Translation Performance](https://aclanthology.org/2025.acl-srw.13/) (Orten et al., ACL 2025)

• Neuron-Level Language Tag Injection Improves Zero-Shot Translation Performance (Orten et al., ACL 2025)

ACL

• Jay Orten, Ammon Shurtz, Nancy Fulda, and Stephen D. Richardson. 2025. Neuron-Level Language Tag Injection Improves Zero-Shot Translation Performance. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 4: Student Research Workshop), pages 203–212, Vienna, Austria. Association for Computational Linguistics.