@inproceedings{tang-etal-2026-diagnosing,
title = "Diagnosing Knowledge Conflict in Multimodal Long-Chain Reasoning",
author = "Tang, Jing and
Wang, Kun and
Lu, Haolang and
Chen, Hongjin and
Chen, KaiTao and
Sun, Zhongxiang and
Li, Qiankun and
Lyu, Lingjuan and
Nan, Guoshun and
Zeng, Zhigang",
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.2240/",
pages = "47947--47986",
ISBN = "979-8-89176-462-0",
abstract = "Multimodal large language models in long chain-of-thought reasoning often fail when different knowledge sources provide conflicting signals. We formalize these failures under a unified notion of knowledge conflict, distinguishing input-level objective conflict from process-level effective conflict. Through probing internal representations, we reveal that: **(I) Linear Separability**: different conflict types are explicitly encoded as linearly separable features rather than entangled; **(II) Depth Localization**: conflict signals concentrate in mid-to-late layers, indicating a distinct processing stage for conflict encoding; **(III) Hierarchical Consistency**: aggregating noisy token-level signals along trajectories robustly recovers input-level conflict types; and **(IV) Directional Asymmetry**: reinforcing the model{'}s implicit source preference under conflict is far easier than enforcing the opposite source. Our findings provide a mechanism-level view of multimodal reasoning under knowledge conflict and enable principled diagnosis and control of long-CoT failures. Code is available at [anonymous link](https://anonymous.4open.science/r/Diagnosing-Knowledge-Conflict-C78D/)."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="tang-etal-2026-diagnosing">
<titleInfo>
<title>Diagnosing Knowledge Conflict in Multimodal Long-Chain Reasoning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jing</namePart>
<namePart type="family">Tang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kun</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haolang</namePart>
<namePart type="family">Lu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hongjin</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">KaiTao</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhongxiang</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qiankun</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lingjuan</namePart>
<namePart type="family">Lyu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guoshun</namePart>
<namePart type="family">Nan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhigang</namePart>
<namePart type="family">Zeng</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>Multimodal large language models in long chain-of-thought reasoning often fail when different knowledge sources provide conflicting signals. We formalize these failures under a unified notion of knowledge conflict, distinguishing input-level objective conflict from process-level effective conflict. Through probing internal representations, we reveal that: **(I) Linear Separability**: different conflict types are explicitly encoded as linearly separable features rather than entangled; **(II) Depth Localization**: conflict signals concentrate in mid-to-late layers, indicating a distinct processing stage for conflict encoding; **(III) Hierarchical Consistency**: aggregating noisy token-level signals along trajectories robustly recovers input-level conflict types; and **(IV) Directional Asymmetry**: reinforcing the model’s implicit source preference under conflict is far easier than enforcing the opposite source. Our findings provide a mechanism-level view of multimodal reasoning under knowledge conflict and enable principled diagnosis and control of long-CoT failures. Code is available at [anonymous link](https://anonymous.4open.science/r/Diagnosing-Knowledge-Conflict-C78D/).</abstract>
<identifier type="citekey">tang-etal-2026-diagnosing</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2240/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>47947</start>
<end>47986</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Diagnosing Knowledge Conflict in Multimodal Long-Chain Reasoning
%A Tang, Jing
%A Wang, Kun
%A Lu, Haolang
%A Chen, Hongjin
%A Chen, KaiTao
%A Sun, Zhongxiang
%A Li, Qiankun
%A Lyu, Lingjuan
%A Nan, Guoshun
%A Zeng, Zhigang
%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 tang-etal-2026-diagnosing
%X Multimodal large language models in long chain-of-thought reasoning often fail when different knowledge sources provide conflicting signals. We formalize these failures under a unified notion of knowledge conflict, distinguishing input-level objective conflict from process-level effective conflict. Through probing internal representations, we reveal that: **(I) Linear Separability**: different conflict types are explicitly encoded as linearly separable features rather than entangled; **(II) Depth Localization**: conflict signals concentrate in mid-to-late layers, indicating a distinct processing stage for conflict encoding; **(III) Hierarchical Consistency**: aggregating noisy token-level signals along trajectories robustly recovers input-level conflict types; and **(IV) Directional Asymmetry**: reinforcing the model’s implicit source preference under conflict is far easier than enforcing the opposite source. Our findings provide a mechanism-level view of multimodal reasoning under knowledge conflict and enable principled diagnosis and control of long-CoT failures. Code is available at [anonymous link](https://anonymous.4open.science/r/Diagnosing-Knowledge-Conflict-C78D/).
%U https://aclanthology.org/2026.findings-emnlp.2240/
%P 47947-47986
Markdown (Informal)
[Diagnosing Knowledge Conflict in Multimodal Long-Chain Reasoning](https://aclanthology.org/2026.findings-emnlp.2240/) (Tang et al., Findings 2026)
ACL
- Jing Tang, Kun Wang, Haolang Lu, Hongjin Chen, KaiTao Chen, Zhongxiang Sun, Qiankun Li, Lingjuan Lyu, Guoshun Nan, and Zhigang Zeng. 2026. Diagnosing Knowledge Conflict in Multimodal Long-Chain Reasoning. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 47947–47986, Budapest, Hungary. Association for Computational Linguistics.