Overcoming Multi-step Complexity in Multimodal Theory-of-Mind Reasoning: A Scalable Bayesian Planner

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Chunhui, Ouyang, Zhongyu, Lee, Kwonjoon, Agarwal, Nakul, Houlihan, Sean Dae, Vosoughi, Soroush, Lo, Shao-Yuan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914546777587712
author Zhang, Chunhui
Ouyang, Zhongyu
Lee, Kwonjoon
Agarwal, Nakul
Houlihan, Sean Dae
Vosoughi, Soroush
Lo, Shao-Yuan
author_facet Zhang, Chunhui
Ouyang, Zhongyu
Lee, Kwonjoon
Agarwal, Nakul
Houlihan, Sean Dae
Vosoughi, Soroush
Lo, Shao-Yuan
contents Theory-of-Mind (ToM) enables humans to infer mental states-such as beliefs, desires, and intentions-forming the foundation of social cognition. However, existing computational ToM methods rely on structured workflows with ToM-specific priors or deep model fine-tuning, which struggle with scalability in multimodal environments and fail to generalize as task complexity increases. To address these limitations, we propose a scalable Bayesian ToM planner that decomposes ToM reasoning into stepwise Bayesian updates. Our framework introduces weak-to-strong control, allowing smaller language models (LMs) to specialize in ToM-specific likelihood estimation and transfer their reasoning behaviors to larger LMs (7B to 405B) for integration with social and world knowledge. This synergistic approach aligns large-model inference of human mental states with Bayesian principles. Extensive experiments show that our method achieves a 4.6% accuracy improvement over state-of-the-art techniques on multimodal ToM benchmarks, including challenging unseen scenarios, thereby establishing a new standard for modeling human mental states in complex environments.
format Preprint
id arxiv_https___arxiv_org_abs_2506_01301
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Overcoming Multi-step Complexity in Multimodal Theory-of-Mind Reasoning: A Scalable Bayesian Planner
Zhang, Chunhui
Ouyang, Zhongyu
Lee, Kwonjoon
Agarwal, Nakul
Houlihan, Sean Dae
Vosoughi, Soroush
Lo, Shao-Yuan
Artificial Intelligence
Computation and Language
Theory-of-Mind (ToM) enables humans to infer mental states-such as beliefs, desires, and intentions-forming the foundation of social cognition. However, existing computational ToM methods rely on structured workflows with ToM-specific priors or deep model fine-tuning, which struggle with scalability in multimodal environments and fail to generalize as task complexity increases. To address these limitations, we propose a scalable Bayesian ToM planner that decomposes ToM reasoning into stepwise Bayesian updates. Our framework introduces weak-to-strong control, allowing smaller language models (LMs) to specialize in ToM-specific likelihood estimation and transfer their reasoning behaviors to larger LMs (7B to 405B) for integration with social and world knowledge. This synergistic approach aligns large-model inference of human mental states with Bayesian principles. Extensive experiments show that our method achieves a 4.6% accuracy improvement over state-of-the-art techniques on multimodal ToM benchmarks, including challenging unseen scenarios, thereby establishing a new standard for modeling human mental states in complex environments.
title Overcoming Multi-step Complexity in Multimodal Theory-of-Mind Reasoning: A Scalable Bayesian Planner
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2506.01301