From Black Boxes to Transparent Minds: Evaluating and Enhancing the Theory of Mind in Multimodal Large Language Models

Xinyang Li*, Siqi Liu*, Bochao Zou†, Jiansheng Chen, Huimin Ma†
University of Science and Technology Beijing
Forty-Second International Conference on Machine Learning 2025
*Indicates Equal Contribution,†Indicates Corresponding Author
Paper Code 🤗 Dataset arXiv
MY ALT TEXT

This illustration highlights the integration of different levels of ToM: recognizing an agent’s desire (Cooper wants to pilot), a first-order belief (he believes he can do it), and a second-order belief (he believes TARS perceives it as risky). These nested mental states are crucial in evaluating advanced ToM.

Abstract

As large language models evolve, there is growing anticipation that they will emulate human-like Theory of Mind (ToM) to assist with routine tasks. However, existing methods for evaluating machine ToM focus primarily on unimodal models and largely treat these models as black boxes, lacking an interpretative exploration of their internal mechanisms. In response, this study adopts an approach based on internal mechanisms to provide an interpretability-driven assessment of ToM in multimodal large language models (MLLMs). Specifically, we first construct a multimodal ToM test dataset, GridToM, which incorporates diverse belief testing tasks and perceptual information from multiple perspectives. Next, our analysis shows that attention heads in multimodal large models can distinguish cognitive information across perspectives, providing evidence of ToM capabilities. Furthermore, we present a lightweight, training-free approach that significantly enhances the model's exhibited ToM by adjusting in the direction of the attention head.

GridToM

For all agents involved in the event, we provide full physical perspective information across the timeline. When an agent closes a door, we mask its perception of any information beyond the door to simulate realistic sensory limitations. Each sample contains three types of questions (illustrated on the right). For each video-text pair, the accompanying text annotations include environment descriptions, initial belief assessments, first-order belief assessments, and second-order belief assessments.

MY ALT TEXT

A FB sample in GridToM includes an omniscient-perspective video covering the entire event timeline, along with omniscient-perspective textual descriptions for each time interval.

Video Examples

Text Examples

Method Overview

MY ALT TEXT

Overview of Our Workflow. We first constructed the GridToM dataset and conducted benchmark testing of MMLMs on it. Subsequently, we input video-text pairs to probe the internal attention representations of the models. Using logistic regression, we performed binary classification on the representations of positive and negative samples to identify attention heads that are sensitive to perspective separation and belief representation. Targeted interventions were then applied to the top K most sensitive attention heads during inference.

Quantitative Results

MY ALT TEXT MY ALT TEXT

BibTeX

@misc{li2025blackboxestransparentminds,
      title={From Black Boxes to Transparent Minds: Evaluating and Enhancing the Theory of Mind in Multimodal Large Language Models}, 
      author={Xinyang Li and Siqi Liu and Bochao Zou and Jiansheng Chen and Huimin Ma},
      year={2025},
      eprint={2506.14224},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2506.14224}, 
}