@inproceedings{shi-cocarascu-2026-end,
title = "End-to-End Argument Mining with {LLM}s: Representation Insights into Multitask Learning",
author = "Shi, Zhiqiang and
Cocarascu, Oana",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.101/",
pages = "2082--2098",
ISBN = "979-8-89176-462-0",
abstract = "Argument mining (AM) traditionally relies on multi-stage pipelines that train separate models for each sub-task, making end-to-end AM challenging and often unstable under multitask learning. In this work, we revisit multitask end-to-end AM in the context of large language models (LLMs) and show that multitask learning behaves fundamentally differently from pre-LLM settings. Across multiple datasets and LLMs, multitask fine-tuning of a single LLM consistently outperforms single-task models, while also exhibiting greater robustness to noisy labels and stronger cross-domain generalization. To understand these qualitatively different behaviors of LLMs, we conduct a layer-wise representation analysis, showing that cross-task features emerge in middle layers, early layers are highly domain-sensitive, and richer middle-layer representations drive performance improvements, robustness, and generalization. We further show that this mechanism extends beyond AM by observing similar patterns across other tasks. Our findings provide new insights into multitask learning with LLMs, highlight its potential for unified, efficient, and robust language understanding, and motivate layer-wise parameter efficient adaptation strategies within and beyond AM."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="shi-cocarascu-2026-end">
<titleInfo>
<title>End-to-End Argument Mining with LLMs: Representation Insights into Multitask Learning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zhiqiang</namePart>
<namePart type="family">Shi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Oana</namePart>
<namePart type="family">Cocarascu</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>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</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">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Argument mining (AM) traditionally relies on multi-stage pipelines that train separate models for each sub-task, making end-to-end AM challenging and often unstable under multitask learning. In this work, we revisit multitask end-to-end AM in the context of large language models (LLMs) and show that multitask learning behaves fundamentally differently from pre-LLM settings. Across multiple datasets and LLMs, multitask fine-tuning of a single LLM consistently outperforms single-task models, while also exhibiting greater robustness to noisy labels and stronger cross-domain generalization. To understand these qualitatively different behaviors of LLMs, we conduct a layer-wise representation analysis, showing that cross-task features emerge in middle layers, early layers are highly domain-sensitive, and richer middle-layer representations drive performance improvements, robustness, and generalization. We further show that this mechanism extends beyond AM by observing similar patterns across other tasks. Our findings provide new insights into multitask learning with LLMs, highlight its potential for unified, efficient, and robust language understanding, and motivate layer-wise parameter efficient adaptation strategies within and beyond AM.</abstract>
<identifier type="citekey">shi-cocarascu-2026-end</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.101/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>2082</start>
<end>2098</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T End-to-End Argument Mining with LLMs: Representation Insights into Multitask Learning
%A Shi, Zhiqiang
%A Cocarascu, Oana
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F shi-cocarascu-2026-end
%X Argument mining (AM) traditionally relies on multi-stage pipelines that train separate models for each sub-task, making end-to-end AM challenging and often unstable under multitask learning. In this work, we revisit multitask end-to-end AM in the context of large language models (LLMs) and show that multitask learning behaves fundamentally differently from pre-LLM settings. Across multiple datasets and LLMs, multitask fine-tuning of a single LLM consistently outperforms single-task models, while also exhibiting greater robustness to noisy labels and stronger cross-domain generalization. To understand these qualitatively different behaviors of LLMs, we conduct a layer-wise representation analysis, showing that cross-task features emerge in middle layers, early layers are highly domain-sensitive, and richer middle-layer representations drive performance improvements, robustness, and generalization. We further show that this mechanism extends beyond AM by observing similar patterns across other tasks. Our findings provide new insights into multitask learning with LLMs, highlight its potential for unified, efficient, and robust language understanding, and motivate layer-wise parameter efficient adaptation strategies within and beyond AM.
%U https://aclanthology.org/2026.findings-emnlp.101/
%P 2082-2098
Markdown (Informal)
[End-to-End Argument Mining with LLMs: Representation Insights into Multitask Learning](https://aclanthology.org/2026.findings-emnlp.101/) (Shi & Cocarascu, Findings 2026)
ACL