@inproceedings{nguyen-salim-2026-whether,
title = "Whether {LLM}s Can Navigate Beliefs and Facts Depends on How You Phrase It",
author = "Nguyen, Quang Minh and
Salim, Luis Frentzen",
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.1171/",
pages = "26862--26872",
ISBN = "979-8-89176-459-0",
abstract = "Humans naturally form and express beliefs in daily communication, e.g., ``I think the answer is 3'' or ``I suppose that{'}s right.'' Such beliefs inevitably intertwine with fact and knowledge, making the ability to handle them in tandem desirable for large language models (LLMs), as they are increasingly deployed in user-facing settings. Prior work showed that even capable LLMs exhibit a systemic weakness in acknowledging user beliefs grounded in incorrect information. We extend this evaluation to 10 LLMs across 18 epistemic expressions and find that the size and direction of this weakness depend on the verb used to express the belief, with the accuracy gap between factual and false information ranging from +50{\%} on ``I vaguely remember'' to -14{\%} on ``I seriously doubt''. We further show that the phenomenon stems from what we call task confusion: models default to fact-checking the underlying claim, overriding the user{'}s stated belief. We provide evidence where chains of thought that explicitly fact-check show lower accuracy on false information than those that do not, and a single instruction can reverse the failure across verb families. Mechanistically, models attend more to false beliefs they fail to confirm, but suppressing this attention at decoding time recovers accuracy only partially and only in some models, calling for future work on intervention methods. Our findings clarify prior results and show how fact-checking, a generally desirable behavior, can interfere with belief tracking in LLMs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="nguyen-salim-2026-whether">
<titleInfo>
<title>Whether LLMs Can Navigate Beliefs and Facts Depends on How You Phrase It</title>
</titleInfo>
<name type="personal">
<namePart type="given">Quang</namePart>
<namePart type="given">Minh</namePart>
<namePart type="family">Nguyen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Luis</namePart>
<namePart type="given">Frentzen</namePart>
<namePart type="family">Salim</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>Humans naturally form and express beliefs in daily communication, e.g., “I think the answer is 3” or “I suppose that’s right.” Such beliefs inevitably intertwine with fact and knowledge, making the ability to handle them in tandem desirable for large language models (LLMs), as they are increasingly deployed in user-facing settings. Prior work showed that even capable LLMs exhibit a systemic weakness in acknowledging user beliefs grounded in incorrect information. We extend this evaluation to 10 LLMs across 18 epistemic expressions and find that the size and direction of this weakness depend on the verb used to express the belief, with the accuracy gap between factual and false information ranging from +50% on “I vaguely remember” to -14% on “I seriously doubt”. We further show that the phenomenon stems from what we call task confusion: models default to fact-checking the underlying claim, overriding the user’s stated belief. We provide evidence where chains of thought that explicitly fact-check show lower accuracy on false information than those that do not, and a single instruction can reverse the failure across verb families. Mechanistically, models attend more to false beliefs they fail to confirm, but suppressing this attention at decoding time recovers accuracy only partially and only in some models, calling for future work on intervention methods. Our findings clarify prior results and show how fact-checking, a generally desirable behavior, can interfere with belief tracking in LLMs.</abstract>
<identifier type="citekey">nguyen-salim-2026-whether</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1171/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>26862</start>
<end>26872</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Whether LLMs Can Navigate Beliefs and Facts Depends on How You Phrase It
%A Nguyen, Quang Minh
%A Salim, Luis Frentzen
%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 nguyen-salim-2026-whether
%X Humans naturally form and express beliefs in daily communication, e.g., “I think the answer is 3” or “I suppose that’s right.” Such beliefs inevitably intertwine with fact and knowledge, making the ability to handle them in tandem desirable for large language models (LLMs), as they are increasingly deployed in user-facing settings. Prior work showed that even capable LLMs exhibit a systemic weakness in acknowledging user beliefs grounded in incorrect information. We extend this evaluation to 10 LLMs across 18 epistemic expressions and find that the size and direction of this weakness depend on the verb used to express the belief, with the accuracy gap between factual and false information ranging from +50% on “I vaguely remember” to -14% on “I seriously doubt”. We further show that the phenomenon stems from what we call task confusion: models default to fact-checking the underlying claim, overriding the user’s stated belief. We provide evidence where chains of thought that explicitly fact-check show lower accuracy on false information than those that do not, and a single instruction can reverse the failure across verb families. Mechanistically, models attend more to false beliefs they fail to confirm, but suppressing this attention at decoding time recovers accuracy only partially and only in some models, calling for future work on intervention methods. Our findings clarify prior results and show how fact-checking, a generally desirable behavior, can interfere with belief tracking in LLMs.
%U https://aclanthology.org/2026.emnlp-main.1171/
%P 26862-26872
Markdown (Informal)
[Whether LLMs Can Navigate Beliefs and Facts Depends on How You Phrase It](https://aclanthology.org/2026.emnlp-main.1171/) (Nguyen & Salim, EMNLP 2026)
ACL