@inproceedings{zhang-etal-2026-cognitive,
title = "Cognitive Analysis Graph-Guided Multi-Turn Safety Enhancement for Large Language Models",
author = "Zhang, Lanxue and
Xie, Yuqiang and
Fang, Fang and
Ren, Yubing and
Wang, Xuebin and
Cao, Yanan",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.1558/",
pages = "31143--31160",
ISBN = "979-8-89176-395-1",
abstract = "Large Language Models exhibit advanced reasoning capabilities that enable them to address complex tasks, but these capabilities also increase the risk of generating harmful content, particularly in multi-turn dialogues. Existing inference-phase safety alignment methods face three major challenges. First, they lack the relationship consideration between question and response, making the model easy to provide harmful content toward complex scenarios. Second, they are difficult to adapt to defense instruction. Third, these methods fail to effectively leverage historical information for safe response generation. To address these challenges, we propose CogGSE, an inference-time safety alignment framework that explicitly models the cognitive process of problem solving through a structured cognitive analysis graph. We retrieve a question-specific graph to ensure the safety information is tailored to the query. To fully exploit historical information in multi-turn settings, we retrieve relevant graphs from previous turns and selectively retain safety-related nodes, which are jointly used with the current-turn graph to guide safe response generation. This design enables transparent, controllable reasoning while maintaining strong safety guarantees. Extensive experiments demonstrate the effectiveness of our approach in multiple safety scenarios."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2026-cognitive">
<titleInfo>
<title>Cognitive Analysis Graph-Guided Multi-Turn Safety Enhancement for Large Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lanxue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuqiang</namePart>
<namePart type="family">Xie</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fang</namePart>
<namePart type="family">Fang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yubing</namePart>
<namePart type="family">Ren</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xuebin</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yanan</namePart>
<namePart type="family">Cao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-395-1</identifier>
</relatedItem>
<abstract>Large Language Models exhibit advanced reasoning capabilities that enable them to address complex tasks, but these capabilities also increase the risk of generating harmful content, particularly in multi-turn dialogues. Existing inference-phase safety alignment methods face three major challenges. First, they lack the relationship consideration between question and response, making the model easy to provide harmful content toward complex scenarios. Second, they are difficult to adapt to defense instruction. Third, these methods fail to effectively leverage historical information for safe response generation. To address these challenges, we propose CogGSE, an inference-time safety alignment framework that explicitly models the cognitive process of problem solving through a structured cognitive analysis graph. We retrieve a question-specific graph to ensure the safety information is tailored to the query. To fully exploit historical information in multi-turn settings, we retrieve relevant graphs from previous turns and selectively retain safety-related nodes, which are jointly used with the current-turn graph to guide safe response generation. This design enables transparent, controllable reasoning while maintaining strong safety guarantees. Extensive experiments demonstrate the effectiveness of our approach in multiple safety scenarios.</abstract>
<identifier type="citekey">zhang-etal-2026-cognitive</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.1558/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>31143</start>
<end>31160</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Cognitive Analysis Graph-Guided Multi-Turn Safety Enhancement for Large Language Models
%A Zhang, Lanxue
%A Xie, Yuqiang
%A Fang, Fang
%A Ren, Yubing
%A Wang, Xuebin
%A Cao, Yanan
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F zhang-etal-2026-cognitive
%X Large Language Models exhibit advanced reasoning capabilities that enable them to address complex tasks, but these capabilities also increase the risk of generating harmful content, particularly in multi-turn dialogues. Existing inference-phase safety alignment methods face three major challenges. First, they lack the relationship consideration between question and response, making the model easy to provide harmful content toward complex scenarios. Second, they are difficult to adapt to defense instruction. Third, these methods fail to effectively leverage historical information for safe response generation. To address these challenges, we propose CogGSE, an inference-time safety alignment framework that explicitly models the cognitive process of problem solving through a structured cognitive analysis graph. We retrieve a question-specific graph to ensure the safety information is tailored to the query. To fully exploit historical information in multi-turn settings, we retrieve relevant graphs from previous turns and selectively retain safety-related nodes, which are jointly used with the current-turn graph to guide safe response generation. This design enables transparent, controllable reasoning while maintaining strong safety guarantees. Extensive experiments demonstrate the effectiveness of our approach in multiple safety scenarios.
%U https://aclanthology.org/2026.findings-acl.1558/
%P 31143-31160
Markdown (Informal)
[Cognitive Analysis Graph-Guided Multi-Turn Safety Enhancement for Large Language Models](https://aclanthology.org/2026.findings-acl.1558/) (Zhang et al., Findings 2026)
ACL