@inproceedings{nallagonda-etal-2026-llm,
title = "{LLM} Summaries Systematically Distort Topic Distributions",
author = "Nallagonda, Shreeya and
Torgbi, Melissa and
Tayyar Madabushi, Harish and
Schockaert, Steven",
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.309/",
pages = "7110--7148",
ISBN = "979-8-89176-459-0",
abstract = "LLM-generated summaries of customer feedback increasingly shape purchasing decisions and organisational strategy. However, we find that such summaries often do not stay faithful to the customer voice, potentially undermining the validity of any downstream decisions made from them. In particular, through a systematic evaluation of 50,000 synthetic customer reviews across 50 domains and 11 models, we demonstrate that the same topics are amplified across different models, regardless of their prominence in the actual reviews. More strikingly, models continue to amplify these same topics even when they are entirely absent from the input, and the topics that are amplified correlate with those the model independently rates as most important. Our findings were validated on real-world reviews from four domains. Furthermore, no intervention we tested, including scaling to frontier models, prompt engineering, and varying temperature, summary length, and input size, reduces the bias to acceptable levels."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="nallagonda-etal-2026-llm">
<titleInfo>
<title>LLM Summaries Systematically Distort Topic Distributions</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shreeya</namePart>
<namePart type="family">Nallagonda</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Melissa</namePart>
<namePart type="family">Torgbi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Harish</namePart>
<namePart type="family">Tayyar Madabushi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Steven</namePart>
<namePart type="family">Schockaert</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>LLM-generated summaries of customer feedback increasingly shape purchasing decisions and organisational strategy. However, we find that such summaries often do not stay faithful to the customer voice, potentially undermining the validity of any downstream decisions made from them. In particular, through a systematic evaluation of 50,000 synthetic customer reviews across 50 domains and 11 models, we demonstrate that the same topics are amplified across different models, regardless of their prominence in the actual reviews. More strikingly, models continue to amplify these same topics even when they are entirely absent from the input, and the topics that are amplified correlate with those the model independently rates as most important. Our findings were validated on real-world reviews from four domains. Furthermore, no intervention we tested, including scaling to frontier models, prompt engineering, and varying temperature, summary length, and input size, reduces the bias to acceptable levels.</abstract>
<identifier type="citekey">nallagonda-etal-2026-llm</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.309/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>7110</start>
<end>7148</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T LLM Summaries Systematically Distort Topic Distributions
%A Nallagonda, Shreeya
%A Torgbi, Melissa
%A Tayyar Madabushi, Harish
%A Schockaert, Steven
%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 nallagonda-etal-2026-llm
%X LLM-generated summaries of customer feedback increasingly shape purchasing decisions and organisational strategy. However, we find that such summaries often do not stay faithful to the customer voice, potentially undermining the validity of any downstream decisions made from them. In particular, through a systematic evaluation of 50,000 synthetic customer reviews across 50 domains and 11 models, we demonstrate that the same topics are amplified across different models, regardless of their prominence in the actual reviews. More strikingly, models continue to amplify these same topics even when they are entirely absent from the input, and the topics that are amplified correlate with those the model independently rates as most important. Our findings were validated on real-world reviews from four domains. Furthermore, no intervention we tested, including scaling to frontier models, prompt engineering, and varying temperature, summary length, and input size, reduces the bias to acceptable levels.
%U https://aclanthology.org/2026.emnlp-main.309/
%P 7110-7148
Markdown (Informal)
[LLM Summaries Systematically Distort Topic Distributions](https://aclanthology.org/2026.emnlp-main.309/) (Nallagonda et al., EMNLP 2026)
ACL
- Shreeya Nallagonda, Melissa Torgbi, Harish Tayyar Madabushi, and Steven Schockaert. 2026. LLM Summaries Systematically Distort Topic Distributions. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 7110–7148, Budapest, Hungary. Association for Computational Linguistics.