@inproceedings{sun-etal-2026-accurate,
title = "Accurate but Not Humble: Evaluating Epistemic Humility in {LLM} Agents under Knowledge Conflict",
author = "Sun, Kaiser and
Jim{\'e}nez Guti{\'e}rrez, Bernal and
Liu, Hongjun and
Zhang, Jingyu and
Gao, Jie and
Dredze, Mark and
Khashabi, Daniel",
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.2128/",
pages = "45401--45420",
ISBN = "979-8-89176-462-0",
abstract = "When retrieved evidence contradicts an agent{'}s prior beliefs, does it revise its answer, acknowledge uncertainty, or persist with an incorrect conclusion? Existing evaluations of agentic systems focus primarily on task success, offering limited insight into how agents handle such conflicts. We propose to evaluate agents on epistemic humility (EH): the agent{'}s willingness to recognize, act on, and communicate uncertainty during task execution. We operationalize EH through three trajectory-level behavioral dimensions: {\_}Identify{\_}, {\_}Solve{\_}, and {\_}Escalate{\_} (ISE). Through {\_}knowledge conflict{\_}, situations where the backbone language model{'}s parametric knowledge contradicts the evidence it encounters, or where two contextual sources disagree, we build a two-level evaluation framework: (1) controlled single-step conflicts, and (2) naturally occurring conflicts during multi-step agentic executions, each level paired with matched no-conflict controls. Evaluating four agents, we find an inverse relationship between task accuracy and epistemic humility: the most accurate agents are often the most likely to give overconfident answers despite recognizing conflicts or risks. Trajectory-level analysis further reveals that agents frequently detect conflicts in early steps of execution but fail to maintain or resolve them in later steps. Finally, we show that Model-level interventions can improve EH, but often at the cost of task accuracy, suggesting that epistemic humility emerges from the interaction among the backbone model, agent harness, and evaluation environment."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="sun-etal-2026-accurate">
<titleInfo>
<title>Accurate but Not Humble: Evaluating Epistemic Humility in LLM Agents under Knowledge Conflict</title>
</titleInfo>
<name type="personal">
<namePart type="given">Kaiser</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bernal</namePart>
<namePart type="family">Jiménez Gutiérrez</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hongjun</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jingyu</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jie</namePart>
<namePart type="family">Gao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mark</namePart>
<namePart type="family">Dredze</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Daniel</namePart>
<namePart type="family">Khashabi</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>When retrieved evidence contradicts an agent’s prior beliefs, does it revise its answer, acknowledge uncertainty, or persist with an incorrect conclusion? Existing evaluations of agentic systems focus primarily on task success, offering limited insight into how agents handle such conflicts. We propose to evaluate agents on epistemic humility (EH): the agent’s willingness to recognize, act on, and communicate uncertainty during task execution. We operationalize EH through three trajectory-level behavioral dimensions: _Identify_, _Solve_, and _Escalate_ (ISE). Through _knowledge conflict_, situations where the backbone language model’s parametric knowledge contradicts the evidence it encounters, or where two contextual sources disagree, we build a two-level evaluation framework: (1) controlled single-step conflicts, and (2) naturally occurring conflicts during multi-step agentic executions, each level paired with matched no-conflict controls. Evaluating four agents, we find an inverse relationship between task accuracy and epistemic humility: the most accurate agents are often the most likely to give overconfident answers despite recognizing conflicts or risks. Trajectory-level analysis further reveals that agents frequently detect conflicts in early steps of execution but fail to maintain or resolve them in later steps. Finally, we show that Model-level interventions can improve EH, but often at the cost of task accuracy, suggesting that epistemic humility emerges from the interaction among the backbone model, agent harness, and evaluation environment.</abstract>
<identifier type="citekey">sun-etal-2026-accurate</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2128/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>45401</start>
<end>45420</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Accurate but Not Humble: Evaluating Epistemic Humility in LLM Agents under Knowledge Conflict
%A Sun, Kaiser
%A Jiménez Gutiérrez, Bernal
%A Liu, Hongjun
%A Zhang, Jingyu
%A Gao, Jie
%A Dredze, Mark
%A Khashabi, Daniel
%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 sun-etal-2026-accurate
%X When retrieved evidence contradicts an agent’s prior beliefs, does it revise its answer, acknowledge uncertainty, or persist with an incorrect conclusion? Existing evaluations of agentic systems focus primarily on task success, offering limited insight into how agents handle such conflicts. We propose to evaluate agents on epistemic humility (EH): the agent’s willingness to recognize, act on, and communicate uncertainty during task execution. We operationalize EH through three trajectory-level behavioral dimensions: _Identify_, _Solve_, and _Escalate_ (ISE). Through _knowledge conflict_, situations where the backbone language model’s parametric knowledge contradicts the evidence it encounters, or where two contextual sources disagree, we build a two-level evaluation framework: (1) controlled single-step conflicts, and (2) naturally occurring conflicts during multi-step agentic executions, each level paired with matched no-conflict controls. Evaluating four agents, we find an inverse relationship between task accuracy and epistemic humility: the most accurate agents are often the most likely to give overconfident answers despite recognizing conflicts or risks. Trajectory-level analysis further reveals that agents frequently detect conflicts in early steps of execution but fail to maintain or resolve them in later steps. Finally, we show that Model-level interventions can improve EH, but often at the cost of task accuracy, suggesting that epistemic humility emerges from the interaction among the backbone model, agent harness, and evaluation environment.
%U https://aclanthology.org/2026.findings-emnlp.2128/
%P 45401-45420
Markdown (Informal)
[Accurate but Not Humble: Evaluating Epistemic Humility in LLM Agents under Knowledge Conflict](https://aclanthology.org/2026.findings-emnlp.2128/) (Sun et al., Findings 2026)
ACL