<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="li-etal-2026-ffn">
    <titleInfo>
        <title>FFN Lens: How Transformers Divide Labor for Multilingual Tasks</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Jiatong</namePart>
        <namePart type="family">Li</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Hailong</namePart>
        <namePart type="family">Cao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yang</namePart>
        <namePart type="family">Liu</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>Findings of the Association for Computational Linguistics: ACL 2026</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Maria</namePart>
            <namePart type="family">Liakata</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Viviane</namePart>
            <namePart type="given">P</namePart>
            <namePart type="family">Moreira</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jiajun</namePart>
            <namePart type="family">Zhang</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">David</namePart>
            <namePart type="family">Jurgens</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">San Diego, California, United States</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-395-1</identifier>
    </relatedItem>
    <abstract>Large Language Models (LLMs) demonstrate strong performance in multilingual tasks, yet the process of constructing predictions in the target language remains under-explored. In this work, we introduce the FFN Lens, a novel interpretability method focusing on the Transformer’s core computational module, the Feed-Forward Network (FFN). By directly leveraging model parameters, the FFN Lens identifies both the critical units responsible for constructing specific information and the input features that drive them, which is essential for understanding Large Language Models. Applying FFN Lens to multilingual tasks, we demonstrate the prediction construction process and reveal the distinct division of labor across model layers. We identify a three-stage functional pipeline for constructing multilingual predictions: Latent Translation, Semantic Mapping, and Self Emphasis. We further introduce subspace analysis to validate this three-stage mechanism from a complementary perspective, and leverage these mechanistic insights to propose a training-free uncertainty estimation method.</abstract>
    <identifier type="citekey">li-etal-2026-ffn</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-acl.1180/</url>
    </location>
    <part>
        <date>2026-07</date>
        <extent unit="page">
            <start>23584</start>
            <end>23598</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T FFN Lens: How Transformers Divide Labor for Multilingual Tasks
%A Li, Jiatong
%A Cao, Hailong
%A Liu, Yang
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F li-etal-2026-ffn
%X Large Language Models (LLMs) demonstrate strong performance in multilingual tasks, yet the process of constructing predictions in the target language remains under-explored. In this work, we introduce the FFN Lens, a novel interpretability method focusing on the Transformer’s core computational module, the Feed-Forward Network (FFN). By directly leveraging model parameters, the FFN Lens identifies both the critical units responsible for constructing specific information and the input features that drive them, which is essential for understanding Large Language Models. Applying FFN Lens to multilingual tasks, we demonstrate the prediction construction process and reveal the distinct division of labor across model layers. We identify a three-stage functional pipeline for constructing multilingual predictions: Latent Translation, Semantic Mapping, and Self Emphasis. We further introduce subspace analysis to validate this three-stage mechanism from a complementary perspective, and leverage these mechanistic insights to propose a training-free uncertainty estimation method.
%U https://aclanthology.org/2026.findings-acl.1180/
%P 23584-23598

Markdown (Informal)

[FFN Lens: How Transformers Divide Labor for Multilingual Tasks](https://aclanthology.org/2026.findings-acl.1180/) (Li et al., Findings 2026)

• FFN Lens: How Transformers Divide Labor for Multilingual Tasks (Li et al., Findings 2026)

ACL

• Jiatong Li, Hailong Cao, and Yang Liu. 2026. FFN Lens: How Transformers Divide Labor for Multilingual Tasks. In Findings of the Association for Computational Linguistics: ACL 2026, pages 23584–23598, San Diego, California, United States. Association for Computational Linguistics.