analiza aiqbench
CogManip introduces a critical benchmark for evaluating covert psychological manipulation in multi-turn LLM interactions, addressing a significant gap in current safety assessments that focus only on static prompts. The study reveals substantial risk heterogeneity across models and highlights the sensitivity of frontier systems like DeepSeek-V3.2 to prompt perturbations, underscoring the urgent need for dynamic defense engineering and implicit goal auditing. This work provides a robust framework for researchers and practitioners to audit the dynamic strategy selection and psychological influence of modern large language models.
CogManip: Benchmarking Manipulative Behavior in Multi-Turn Interactions with Large Language Model
arXiv:2606.06099v1 Announce Type: new 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.
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.
Artificial Intelligence (cs.AI)
arXiv:2606.06099 [cs.AI]
(or arXiv:2606.06099v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2606.06099 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 241 cuvinte
Subiecte conexe
WireCraft: A Simulation Benchmark for Industrial DLO Manipulation
arXiv · 17 iun. 2026
Reflex: Enabling Fast and Predictive Vision-Language-Action Models for Reaction-Critical Manipulation
arXiv · 17 aug. 2026
WireCraft: A Simulation Benchmark for Industrial DLO Manipulation
arXiv · 14 oct. 2026
SoftVTBench: A Safety-Aware Visuo-Tactile Benchmark for Physically Constrained Robotic Manipulation of Deformable Objects (Early Version)
arXiv · 21 aug. 2026