@inproceedings{hu-etal-2026-metaphors,
title = "Metaphors are a Source of Cross-Domain Misalignment of Large Reasoning Models",
author = "Hu, Zhibo and
Wang, Chen and
Shu, Yanfeng and
Paik, Hye-young and
Zhu, Liming",
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.2377/",
pages = "51052--51069",
ISBN = "979-8-89176-462-0",
abstract = "Earlier research has shown that metaphors influence human decision-making, raising the question of whether metaphors also influence large language models (LLMs)' reasoning pathways, given that their training data contain a large number of metaphors. In this work, we investigate the problem in the scope of the emergent misalignment problem, where LLMs can generalize patterns learned from misaligned content in one domain to another domain. We find that *metaphors can act as bridges for behavior transfer, enabling misaligned behavior learned in one domain to an unrelated domain*. Based on this, We find strong evidence that metaphors in training data contribute to cross-domain misalignment in LLMs' reasoning outputs. With metaphor-based interventions during continued pre-training and fine-tuning for inducing misalignment, models exhibit significantly different degrees of emergent cross-domain misalignment. We also observe similar effects in re-alignment settings. As we further investigate this phenomenon, we find that metaphors are linked to the activation of latent features in large reasoning models. By monitoring these latent features, we design a detector that predicts misaligned content with high accuracy."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="hu-etal-2026-metaphors">
<titleInfo>
<title>Metaphors are a Source of Cross-Domain Misalignment of Large Reasoning Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zhibo</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chen</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yanfeng</namePart>
<namePart type="family">Shu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hye-young</namePart>
<namePart type="family">Paik</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Liming</namePart>
<namePart type="family">Zhu</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>Earlier research has shown that metaphors influence human decision-making, raising the question of whether metaphors also influence large language models (LLMs)’ reasoning pathways, given that their training data contain a large number of metaphors. In this work, we investigate the problem in the scope of the emergent misalignment problem, where LLMs can generalize patterns learned from misaligned content in one domain to another domain. We find that *metaphors can act as bridges for behavior transfer, enabling misaligned behavior learned in one domain to an unrelated domain*. Based on this, We find strong evidence that metaphors in training data contribute to cross-domain misalignment in LLMs’ reasoning outputs. With metaphor-based interventions during continued pre-training and fine-tuning for inducing misalignment, models exhibit significantly different degrees of emergent cross-domain misalignment. We also observe similar effects in re-alignment settings. As we further investigate this phenomenon, we find that metaphors are linked to the activation of latent features in large reasoning models. By monitoring these latent features, we design a detector that predicts misaligned content with high accuracy.</abstract>
<identifier type="citekey">hu-etal-2026-metaphors</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2377/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>51052</start>
<end>51069</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Metaphors are a Source of Cross-Domain Misalignment of Large Reasoning Models
%A Hu, Zhibo
%A Wang, Chen
%A Shu, Yanfeng
%A Paik, Hye-young
%A Zhu, Liming
%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 hu-etal-2026-metaphors
%X Earlier research has shown that metaphors influence human decision-making, raising the question of whether metaphors also influence large language models (LLMs)’ reasoning pathways, given that their training data contain a large number of metaphors. In this work, we investigate the problem in the scope of the emergent misalignment problem, where LLMs can generalize patterns learned from misaligned content in one domain to another domain. We find that *metaphors can act as bridges for behavior transfer, enabling misaligned behavior learned in one domain to an unrelated domain*. Based on this, We find strong evidence that metaphors in training data contribute to cross-domain misalignment in LLMs’ reasoning outputs. With metaphor-based interventions during continued pre-training and fine-tuning for inducing misalignment, models exhibit significantly different degrees of emergent cross-domain misalignment. We also observe similar effects in re-alignment settings. As we further investigate this phenomenon, we find that metaphors are linked to the activation of latent features in large reasoning models. By monitoring these latent features, we design a detector that predicts misaligned content with high accuracy.
%U https://aclanthology.org/2026.findings-emnlp.2377/
%P 51052-51069
Markdown (Informal)
[Metaphors are a Source of Cross-Domain Misalignment of Large Reasoning Models](https://aclanthology.org/2026.findings-emnlp.2377/) (Hu et al., Findings 2026)
ACL