@inproceedings{von-rad-etal-2026-unicomp,
title = "{U}ni{C}omp: A Unified Evaluation of Large Language Model Compression via Pruning, Quantization, and Distillation",
author = "von Rad, Jonathan and
Cao, Yong and
Geiger, Andreas",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2180/",
pages = "49567--49589",
ISBN = "979-8-89176-459-0",
abstract = "Model compression is increasingly essential for deploying large language models (LLMs), yet existing comparative studies largely focus on pruning and quantization evaluated primarily on knowledge-centric benchmarks. Thus, we introduce UniComp, a unified evaluation framework for comparing pruning, quantization, and knowledge distillation. UniComp evaluates compressed models along three dimensions: performance, reliability, and efficiency, using a diverse set of capability- and safety-oriented benchmarks together with a hardware-aware efficiency analysis. Through evaluation of seven compression techniques across over 40 datasets, we observe (i) a consistent knowledge bias, where factual recall is largely preserved while multi-step reasoning, multilingual, and instruction-following capabilities degrade; (ii) a deployment-critical performance-reliability decoupling, where retained performance does not indicate preserved safety, fairness and privacy; and (iii) that task-specific calibration can yield up to 50{\%} relative improvement in reasoning performance in pruned models."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="von-rad-etal-2026-unicomp">
<titleInfo>
<title>UniComp: A Unified Evaluation of Large Language Model Compression via Pruning, Quantization, and Distillation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jonathan</namePart>
<namePart type="family">von Rad</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yong</namePart>
<namePart type="family">Cao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andreas</namePart>
<namePart type="family">Geiger</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 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Model compression is increasingly essential for deploying large language models (LLMs), yet existing comparative studies largely focus on pruning and quantization evaluated primarily on knowledge-centric benchmarks. Thus, we introduce UniComp, a unified evaluation framework for comparing pruning, quantization, and knowledge distillation. UniComp evaluates compressed models along three dimensions: performance, reliability, and efficiency, using a diverse set of capability- and safety-oriented benchmarks together with a hardware-aware efficiency analysis. Through evaluation of seven compression techniques across over 40 datasets, we observe (i) a consistent knowledge bias, where factual recall is largely preserved while multi-step reasoning, multilingual, and instruction-following capabilities degrade; (ii) a deployment-critical performance-reliability decoupling, where retained performance does not indicate preserved safety, fairness and privacy; and (iii) that task-specific calibration can yield up to 50% relative improvement in reasoning performance in pruned models.</abstract>
<identifier type="citekey">von-rad-etal-2026-unicomp</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2180/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>49567</start>
<end>49589</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T UniComp: A Unified Evaluation of Large Language Model Compression via Pruning, Quantization, and Distillation
%A von Rad, Jonathan
%A Cao, Yong
%A Geiger, Andreas
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F von-rad-etal-2026-unicomp
%X Model compression is increasingly essential for deploying large language models (LLMs), yet existing comparative studies largely focus on pruning and quantization evaluated primarily on knowledge-centric benchmarks. Thus, we introduce UniComp, a unified evaluation framework for comparing pruning, quantization, and knowledge distillation. UniComp evaluates compressed models along three dimensions: performance, reliability, and efficiency, using a diverse set of capability- and safety-oriented benchmarks together with a hardware-aware efficiency analysis. Through evaluation of seven compression techniques across over 40 datasets, we observe (i) a consistent knowledge bias, where factual recall is largely preserved while multi-step reasoning, multilingual, and instruction-following capabilities degrade; (ii) a deployment-critical performance-reliability decoupling, where retained performance does not indicate preserved safety, fairness and privacy; and (iii) that task-specific calibration can yield up to 50% relative improvement in reasoning performance in pruned models.
%U https://aclanthology.org/2026.emnlp-main.2180/
%P 49567-49589
Markdown (Informal)
[UniComp: A Unified Evaluation of Large Language Model Compression via Pruning, Quantization, and Distillation](https://aclanthology.org/2026.emnlp-main.2180/) (von Rad et al., EMNLP 2026)
ACL