Introduces NCP-ExploreToM framework to evaluate LLMs on inducing belief states via planning and action, with GPT-5 succeeding on ~80% of tasks and outperforming humans.
arXiv preprint arXiv:2502.11881 , year=
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
representative citing papers
Tool-using LLM agents can implement undetectable stegosystems, shifting the primary barrier to covert multi-agent collusion from technical feasibility to coordination without explicit agreement.
Improvements in LLM Theory of Mind on static benchmarks do not reliably improve performance in dynamic, first-person human-AI interactions across goal-oriented and experience-oriented tasks.
UserHarness reframes ToM as explicit user-mind reconstruction and reports up to 95.94% macro accuracy on five benchmarks with over 15% relative gains.
A survey proposing a three-level capability taxonomy (L1 Predictor, L2 Simulator, L3 Evolver) for world models across physical, digital, social, and scientific domains.
Observational analysis of Brazilian YouTube climate content identifies psychological engagement traits and explores their use in generative AI campaigns, accompanied by a public dataset of 226K videos and 2.7M comments.
citing papers explorer
-
Theory of Mind and Persuasion Beyond Conversation: Assessing the Capacity of LLMs to Induce Belief States via Planning and Action
Introduces NCP-ExploreToM framework to evaluate LLMs on inducing belief states via planning and action, with GPT-5 succeeding on ~80% of tasks and outperforming humans.
-
Tool Use Enables Undetectable Steganography in Multi-Agent LLM Systems
Tool-using LLM agents can implement undetectable stegosystems, shifting the primary barrier to covert multi-agent collusion from technical feasibility to coordination without explicit agreement.
-
Does Theory of Mind Improvement Really Benefit Human-AI Interactions? Empirical Findings from Interactive Evaluations
Improvements in LLM Theory of Mind on static benchmarks do not reliably improve performance in dynamic, first-person human-AI interactions across goal-oriented and experience-oriented tasks.
-
UserHarness: Harnessing User Minds for Stronger Agent Theory-of-Mind
UserHarness reframes ToM as explicit user-mind reconstruction and reports up to 95.94% macro accuracy on five benchmarks with over 15% relative gains.
-
Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond
A survey proposing a three-level capability taxonomy (L1 Predictor, L2 Simulator, L3 Evolver) for world models across physical, digital, social, and scientific domains.
-
Characterizing AI Manipulation Risks in Brazilian YouTube Climate Discourse
Observational analysis of Brazilian YouTube climate content identifies psychological engagement traits and explores their use in generative AI campaigns, accompanied by a public dataset of 226K videos and 2.7M comments.