@inproceedings{gampel-etal-2026-need,
title = "Do We Need Large Models for Argument Classification? Revisiting the Role of Model Compression",
author = "Gampel, Filip and
Olszowski, Rafa{\l} and
Pietro{\'n}, Marcin",
editor = "Elaraby, Mohamed and
Hautli-Janisz, Annette and
Romberg, Julia and
Musi, Elena and
Ruggeri, Federico and
Lawrence, John",
booktitle = "Proceedings of the 13th Workshop on Argument Mining and Reasoning",
month = jul,
year = "2026",
address = "San Diego, California, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.argmining-1.8/",
pages = "66--73",
ISBN = "979-8-89176-399-9",
abstract = "Large language models have improved argument mining substantially, but the associated computational cost complicates deployment, replication, and systematic comparison. We examine how much compression an open-source large language model can tolerate before argument classification quality degrades. Using gpt-oss-20b as the base model, we study pruning with Wanda and post-training quantization under a zero-shot prompting setup. We evaluate compressed variants on three argument-mining resources, namely UKP, Args.me, and ARIES, and contrast their behavior with general language-model benchmarks. The results show a consistent pattern: moderate pruning preserves most of the original performance on argument classification, whereas activation quantization causes larger and more systematic drops. The findings suggest that argument classification is more compression-tolerant than general-purpose evaluation suites, but only up to a point, and they should not be interpreted as evidence that aggressive compression is universally safe. We therefore position compression as a practical way to reduce model cost for argument analysis, while emphasizing that claims about efficiency gains must distinguish between preserved predictive quality and realized runtime speedups."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="gampel-etal-2026-need">
<titleInfo>
<title>Do We Need Large Models for Argument Classification? Revisiting the Role of Model Compression</title>
</titleInfo>
<name type="personal">
<namePart type="given">Filip</namePart>
<namePart type="family">Gampel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rafał</namePart>
<namePart type="family">Olszowski</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marcin</namePart>
<namePart type="family">Pietroń</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>Proceedings of the 13th Workshop on Argument Mining and Reasoning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mohamed</namePart>
<namePart type="family">Elaraby</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Annette</namePart>
<namePart type="family">Hautli-Janisz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Julia</namePart>
<namePart type="family">Romberg</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Elena</namePart>
<namePart type="family">Musi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Federico</namePart>
<namePart type="family">Ruggeri</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">John</namePart>
<namePart type="family">Lawrence</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, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-399-9</identifier>
</relatedItem>
<abstract>Large language models have improved argument mining substantially, but the associated computational cost complicates deployment, replication, and systematic comparison. We examine how much compression an open-source large language model can tolerate before argument classification quality degrades. Using gpt-oss-20b as the base model, we study pruning with Wanda and post-training quantization under a zero-shot prompting setup. We evaluate compressed variants on three argument-mining resources, namely UKP, Args.me, and ARIES, and contrast their behavior with general language-model benchmarks. The results show a consistent pattern: moderate pruning preserves most of the original performance on argument classification, whereas activation quantization causes larger and more systematic drops. The findings suggest that argument classification is more compression-tolerant than general-purpose evaluation suites, but only up to a point, and they should not be interpreted as evidence that aggressive compression is universally safe. We therefore position compression as a practical way to reduce model cost for argument analysis, while emphasizing that claims about efficiency gains must distinguish between preserved predictive quality and realized runtime speedups.</abstract>
<identifier type="citekey">gampel-etal-2026-need</identifier>
<location>
<url>https://aclanthology.org/2026.argmining-1.8/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>66</start>
<end>73</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Do We Need Large Models for Argument Classification? Revisiting the Role of Model Compression
%A Gampel, Filip
%A Olszowski, Rafał
%A Pietroń, Marcin
%Y Elaraby, Mohamed
%Y Hautli-Janisz, Annette
%Y Romberg, Julia
%Y Musi, Elena
%Y Ruggeri, Federico
%Y Lawrence, John
%S Proceedings of the 13th Workshop on Argument Mining and Reasoning
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, USA
%@ 979-8-89176-399-9
%F gampel-etal-2026-need
%X Large language models have improved argument mining substantially, but the associated computational cost complicates deployment, replication, and systematic comparison. We examine how much compression an open-source large language model can tolerate before argument classification quality degrades. Using gpt-oss-20b as the base model, we study pruning with Wanda and post-training quantization under a zero-shot prompting setup. We evaluate compressed variants on three argument-mining resources, namely UKP, Args.me, and ARIES, and contrast their behavior with general language-model benchmarks. The results show a consistent pattern: moderate pruning preserves most of the original performance on argument classification, whereas activation quantization causes larger and more systematic drops. The findings suggest that argument classification is more compression-tolerant than general-purpose evaluation suites, but only up to a point, and they should not be interpreted as evidence that aggressive compression is universally safe. We therefore position compression as a practical way to reduce model cost for argument analysis, while emphasizing that claims about efficiency gains must distinguish between preserved predictive quality and realized runtime speedups.
%U https://aclanthology.org/2026.argmining-1.8/
%P 66-73
Markdown (Informal)
[Do We Need Large Models for Argument Classification? Revisiting the Role of Model Compression](https://aclanthology.org/2026.argmining-1.8/) (Gampel et al., ArgMining 2026)
ACL