AnnaLeee/GridToM

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
*Indicates Equal Contribution,†Indicates Corresponding Author

🏆 Overview

This repository provides the code for the paper [ICML 2025] "From Black Boxes to Transparent Minds: Evaluating and Enhancing the Theory of Mind in Multimodal Large Language Models"

⚙️ Installation

conda create -n gridtom python=3.12
conda activate gridtom
# Please install PyTorch according to your CUDA version.
conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia
pip install -r requirements.txt

⚡️ Quick Start

chmod 700 *.sh

./init_belief.sh LLaVA-NeXT-Video-7B-hf
./init_belief.sh Qwen2-VL-7B-Instruct

./evaluate.sh LLaVA-NeXT-Video-7B-hf
./evaluate.sh Qwen2-VL-7B-Instruct

./save_states.sh LLaVA-NeXT-Video-7B-hf
./save_states.sh Qwen2-VL-7B-Instruct

./interv_evaluate.sh LLaVA-NeXT-Video-7B-hf
./interv_evaluate.sh Qwen2-VL-7B-Instruct

🔎 Citation

Please cite the paper and star this repo if you find it interesting/useful, thanks!

@article{li2025black,
  title={From Black Boxes to Transparent Minds: Evaluating and Enhancing the Theory of Mind in Multimodal Large Language Models},
  author={Li, Xinyang and Liu, Siqi and Zou, Bochao and Chen, Jiansheng and Ma, Huimin},
  journal={arXiv preprint arXiv:2506.14224},
  year={2025}
}

🌟 Star History