Hypothesis-Driven Theory-of-Mind Reasoning for Large Language Models

2025-02-17

AI Summary

Large language models (LLMs) excel at deterministic tasks such as mathematics and programming but remain limited in social reasoning—specifically, dynamically inferring others’ latent mental states (e.g., beliefs, intentions)—due to the absence of ground-truth annotations and formal verification criteria. To address this, we propose *Mind Tracking*, the first algorithm that integrates Bayesian theory of mind with sequential Monte Carlo methods to enable unsupervised, observation-driven mental-state modeling. Mind Tracking operates by generating hypotheses about agents’ hidden mental states, dynamically weighting them via likelihood-based evidence, and recursively updating belief distributions in real time. Evaluated across multiple theory-of-mind benchmarks, it significantly outperforms state-of-the-art LLMs—including o1 and R1—and uncovers distinct behavioral patterns in their social reasoning. Our approach endows LLMs with robust, interpretable, and computationally grounded mental-state tracking capabilities, advancing principled social cognition in foundation models.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Knowledge Representation and Reasoning: Reasoning with Beliefs

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Existing LLM reasoning methods have shown impressive capabilities across various tasks, such as solving math and coding problems. However, applying these methods to scenarios without ground-truth answers or rule-based verification methods - such as tracking the mental states of an agent - remains challenging. Inspired by the sequential Monte Carlo algorithm, we introduce thought-tracing, an inference-time reasoning algorithm designed to trace the mental states of specific agents by generating hypotheses and weighting them based on observations without relying on ground-truth solutions to questions in datasets. Our algorithm is modeled after the Bayesian theory-of-mind framework, using LLMs to approximate probabilistic inference over agents' evolving mental states based on their perceptions and actions. We evaluate thought-tracing on diverse theory-of-mind benchmarks, demonstrating significant performance improvements compared to baseline LLMs. Our experiments also reveal interesting behaviors of the recent reasoning models - e.g., o1 and R1 - on theory-of-mind, highlighting the difference of social reasoning compared to other domains.

Problem

• Tracking mental states of agents
• Hypothesis-driven reasoning without ground-truth
• Bayesian theory-of-mind framework for LLMs

Innovation

• Thought-tracing algorithm introduced
• Bayesian theory-of-mind framework used
• LLMs approximate probabilistic inference