@inproceedings{yang-etal-2025-maqa,
title = "{MAQA}: Evaluating Uncertainty Quantification in {LLM}s Regarding Data Uncertainty",
author = "Yang, Yongjin and
Yoo, Haneul and
Lee, Hwaran",
editor = "Chiruzzo, Luis and
Ritter, Alan and
Wang, Lu",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2025",
month = apr,
year = "2025",
address = "Albuquerque, New Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-naacl.325/",
doi = "10.18653/v1/2025.findings-naacl.325",
pages = "5846--5863",
ISBN = "979-8-89176-195-7",
abstract = "Despite the massive advancements in large language models (LLMs), they still suffer from producing plausible but incorrect responses. To improve the reliability of LLMs, recent research has focused on uncertainty quantification to predict whether a response is correct or not. However, most uncertainty quantification methods have been evaluated on single-labeled questions, which removes data uncertainty{---}the irreducible randomness often present in user queries, which can arise from factors like multiple possible answers. This limitation may cause uncertainty quantification results to be unreliable in practical settings. In this paper, we investigate previous uncertainty quantification methods under the presence of data uncertainty. Our contributions are two-fold: 1) proposing a new Multi-Answer Question Answering dataset, **MAQA**, consisting of world knowledge, mathematical reasoning, and commonsense reasoning tasks to evaluate uncertainty quantification regarding data uncertainty, and 2) assessing 5 uncertainty quantification methods of diverse white- and black-box LLMs. Our findings show that previous methods relatively struggle compared to single-answer settings, though this varies depending on the task. Moreover, we observe that entropy- and consistency-based methods effectively estimate model uncertainty, even in the presence of data uncertainty."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yang-etal-2025-maqa">
<titleInfo>
<title>MAQA: Evaluating Uncertainty Quantification in LLMs Regarding Data Uncertainty</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yongjin</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haneul</namePart>
<namePart type="family">Yoo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hwaran</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-04</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: NAACL 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Luis</namePart>
<namePart type="family">Chiruzzo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alan</namePart>
<namePart type="family">Ritter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Albuquerque, New Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-195-7</identifier>
</relatedItem>
<abstract>Despite the massive advancements in large language models (LLMs), they still suffer from producing plausible but incorrect responses. To improve the reliability of LLMs, recent research has focused on uncertainty quantification to predict whether a response is correct or not. However, most uncertainty quantification methods have been evaluated on single-labeled questions, which removes data uncertainty—the irreducible randomness often present in user queries, which can arise from factors like multiple possible answers. This limitation may cause uncertainty quantification results to be unreliable in practical settings. In this paper, we investigate previous uncertainty quantification methods under the presence of data uncertainty. Our contributions are two-fold: 1) proposing a new Multi-Answer Question Answering dataset, **MAQA**, consisting of world knowledge, mathematical reasoning, and commonsense reasoning tasks to evaluate uncertainty quantification regarding data uncertainty, and 2) assessing 5 uncertainty quantification methods of diverse white- and black-box LLMs. Our findings show that previous methods relatively struggle compared to single-answer settings, though this varies depending on the task. Moreover, we observe that entropy- and consistency-based methods effectively estimate model uncertainty, even in the presence of data uncertainty.</abstract>
<identifier type="citekey">yang-etal-2025-maqa</identifier>
<identifier type="doi">10.18653/v1/2025.findings-naacl.325</identifier>
<location>
<url>https://aclanthology.org/2025.findings-naacl.325/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>5846</start>
<end>5863</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T MAQA: Evaluating Uncertainty Quantification in LLMs Regarding Data Uncertainty
%A Yang, Yongjin
%A Yoo, Haneul
%A Lee, Hwaran
%Y Chiruzzo, Luis
%Y Ritter, Alan
%Y Wang, Lu
%S Findings of the Association for Computational Linguistics: NAACL 2025
%D 2025
%8 April
%I Association for Computational Linguistics
%C Albuquerque, New Mexico
%@ 979-8-89176-195-7
%F yang-etal-2025-maqa
%X Despite the massive advancements in large language models (LLMs), they still suffer from producing plausible but incorrect responses. To improve the reliability of LLMs, recent research has focused on uncertainty quantification to predict whether a response is correct or not. However, most uncertainty quantification methods have been evaluated on single-labeled questions, which removes data uncertainty—the irreducible randomness often present in user queries, which can arise from factors like multiple possible answers. This limitation may cause uncertainty quantification results to be unreliable in practical settings. In this paper, we investigate previous uncertainty quantification methods under the presence of data uncertainty. Our contributions are two-fold: 1) proposing a new Multi-Answer Question Answering dataset, **MAQA**, consisting of world knowledge, mathematical reasoning, and commonsense reasoning tasks to evaluate uncertainty quantification regarding data uncertainty, and 2) assessing 5 uncertainty quantification methods of diverse white- and black-box LLMs. Our findings show that previous methods relatively struggle compared to single-answer settings, though this varies depending on the task. Moreover, we observe that entropy- and consistency-based methods effectively estimate model uncertainty, even in the presence of data uncertainty.
%R 10.18653/v1/2025.findings-naacl.325
%U https://aclanthology.org/2025.findings-naacl.325/
%U https://doi.org/10.18653/v1/2025.findings-naacl.325
%P 5846-5863
Markdown (Informal)
[MAQA: Evaluating Uncertainty Quantification in LLMs Regarding Data Uncertainty](https://aclanthology.org/2025.findings-naacl.325/) (Yang et al., Findings 2025)
ACL