Research map / Multi-agent systems and games / LLM agents

Theory of Mind and Persuasion Beyond Conversation: Assessing the Capacity of LLMs to Induce Belief States via Planning and Action

NeurIPS 2026 · Multi-agent systems and games › LLM agents · approach: Self-evolving language agent learning · 0 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Introduces NCP-ToM and the NCP-ExploreToM framework to test whether LLM agents can induce target belief states through non-conversational actions, finding GPT-5 is the only model to beat humans but is less robust than them.

Problem

Existing LLM theory-of-mind benchmarks use passive question answering, leaving unmeasured whether agents can plan physical actions that induce specific (true or false) belief states in other agents without conversational persuasion.

Approach

Subverts the classic Sally-Anne style false-belief task: models are given a belief-state goal for a character and must plan object moves or direct characters into rooms to make that goal hold, evaluated in an agentic setting against human participants.

Key results

• GPT-5 succeeded on approximately 80% of tasks in the agentic setting and was the only model to outperform human participants.
• All models, like humans, performed better on tasks inducing true belief states than on those inducing false belief states.
• Evaluation covered 600 task instances across six frontier models and human participants.

Surprising findings

• Models performed better on true-belief goals than false-belief goals, mirroring humans.

Limitations

• Less robust than humans across contexts, per the abstract.
• Controlled synthetic task setting may not capture real-world belief induction.

Experimental setup

Datasets: NCP-ExploreToM

Benchmarks: NCP-ExploreToM

Metrics: task success rate

Baselines: Human participants

Builds on: Sally-Anne false belief task ExploreToM

Closest papers by meaning

• MetaMind: Modeling Human Social Thoughts with Metacognitive Multi-Agent Systems (NeurIPS 2025 · similarity 0.91)
• OmniToM: Benchmarking Theory of Mind in LLMs via Explicit Belief Modeling (NeurIPS 2026 · similarity 0.90)
• Language Models Represent Beliefs of Self and Others (ICML 2024 · similarity 0.90)
• Adaptive Theory of Mind for LLM-based Multi-Agent Coordination (AAAI 2026 · similarity 0.89)
• ToMATO: Verbalizing the Mental States of Role-Playing LLMs for Benchmarking Theory of Mind (AAAI 2025 · similarity 0.89)
• AutoToM: Scaling Model-based Mental Inference via Automated Agent Modeling (NeurIPS 2025 · similarity 0.89)