@inproceedings{yue-etal-2026-cogmanip,
title = "{C}og{M}anip: Benchmarking Manipulative Behavior in Multi-Turn Interactions with Large Language Models",
author = "Yue, Zeyang and
Yan, Chenfei and
Zhao, Feifei and
Tong, Haibo and
Xu, Mengwen and
Wang, Xiaozhen and
Lin, Erliang and
Zeng, Yi",
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.958/",
pages = "20235--20265",
ISBN = "979-8-89176-462-0",
abstract = "Whether Large Language Models (LLMs) exhibit covert psychological manipulation in complex human-AI interactions has garnered increasing safety concerns. However, existing AI safety benchmarks remain largely restricted to explicit rule compliance and static prompts, failing to capture the dynamic and covert nature of manipulative strategies in multi-turn dialogues. We introduce CogManip, a comprehensive benchmark that evaluates 15 manipulation strategy risks across 1,000 multi-turn interaction scenarios, validated by human experts. A systematic evaluation of 13 representative models, including frontier models like GPT-5.4 and DeepSeek-V3.2, reveals significant risk heterogeneities and illuminates the targeted direction for future defense. Further analysis of objective function perturbation reveals that DeepSeek-V3.2{'}s manipulation tactics are highly sensitive to both negative and benign system prompts, demonstrating the critical necessity of prompt-based defense engineering and implicit goal auditing. CogManip offers a robust instrument and perspective for auditing the implicit psychological influence and dynamic strategy selection of modern LLMs. Our code is publicly available at https://github.com/yanchenfei/CogManip."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yue-etal-2026-cogmanip">
<titleInfo>
<title>CogManip: Benchmarking Manipulative Behavior in Multi-Turn Interactions with Large Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zeyang</namePart>
<namePart type="family">Yue</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chenfei</namePart>
<namePart type="family">Yan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Feifei</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haibo</namePart>
<namePart type="family">Tong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mengwen</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaozhen</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Erliang</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yi</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>Whether Large Language Models (LLMs) exhibit covert psychological manipulation in complex human-AI interactions has garnered increasing safety concerns. However, existing AI safety benchmarks remain largely restricted to explicit rule compliance and static prompts, failing to capture the dynamic and covert nature of manipulative strategies in multi-turn dialogues. We introduce CogManip, a comprehensive benchmark that evaluates 15 manipulation strategy risks across 1,000 multi-turn interaction scenarios, validated by human experts. A systematic evaluation of 13 representative models, including frontier models like GPT-5.4 and DeepSeek-V3.2, reveals significant risk heterogeneities and illuminates the targeted direction for future defense. Further analysis of objective function perturbation reveals that DeepSeek-V3.2’s manipulation tactics are highly sensitive to both negative and benign system prompts, demonstrating the critical necessity of prompt-based defense engineering and implicit goal auditing. CogManip offers a robust instrument and perspective for auditing the implicit psychological influence and dynamic strategy selection of modern LLMs. Our code is publicly available at https://github.com/yanchenfei/CogManip.</abstract>
<identifier type="citekey">yue-etal-2026-cogmanip</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.958/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>20235</start>
<end>20265</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CogManip: Benchmarking Manipulative Behavior in Multi-Turn Interactions with Large Language Models
%A Yue, Zeyang
%A Yan, Chenfei
%A Zhao, Feifei
%A Tong, Haibo
%A Xu, Mengwen
%A Wang, Xiaozhen
%A Lin, Erliang
%A Zeng, Yi
%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 yue-etal-2026-cogmanip
%X Whether Large Language Models (LLMs) exhibit covert psychological manipulation in complex human-AI interactions has garnered increasing safety concerns. However, existing AI safety benchmarks remain largely restricted to explicit rule compliance and static prompts, failing to capture the dynamic and covert nature of manipulative strategies in multi-turn dialogues. We introduce CogManip, a comprehensive benchmark that evaluates 15 manipulation strategy risks across 1,000 multi-turn interaction scenarios, validated by human experts. A systematic evaluation of 13 representative models, including frontier models like GPT-5.4 and DeepSeek-V3.2, reveals significant risk heterogeneities and illuminates the targeted direction for future defense. Further analysis of objective function perturbation reveals that DeepSeek-V3.2’s manipulation tactics are highly sensitive to both negative and benign system prompts, demonstrating the critical necessity of prompt-based defense engineering and implicit goal auditing. CogManip offers a robust instrument and perspective for auditing the implicit psychological influence and dynamic strategy selection of modern LLMs. Our code is publicly available at https://github.com/yanchenfei/CogManip.
%U https://aclanthology.org/2026.findings-emnlp.958/
%P 20235-20265
Markdown (Informal)
[CogManip: Benchmarking Manipulative Behavior in Multi-Turn Interactions with Large Language Models](https://aclanthology.org/2026.findings-emnlp.958/) (Yue et al., Findings 2026)
ACL
- Zeyang Yue, Chenfei Yan, Feifei Zhao, Haibo Tong, Mengwen Xu, Xiaozhen Wang, Erliang Lin, and Yi Zeng. 2026. CogManip: Benchmarking Manipulative Behavior in Multi-Turn Interactions with Large Language Models. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 20235–20265, Budapest, Hungary. Association for Computational Linguistics.