As Large Language Models (LLMs) increasingly participate in human-AI interactions, evaluating their Theory of Mind (ToM) capabilities - particularly their ability to track dynamic mental states - becomes crucial. While existing benchmarks assess basic ToM abilities, they predominantly focus on static snapshots of mental states, overlooking the temporal evolution that characterizes real-world social interactions. We present \textsc{DynToM}, a novel benchmark specifically designed to evaluate LLMs' ability to understand and track the temporal progression of mental states across interconnected scenarios. Through a systematic four-step framework, we generate 1,100 social contexts encompassing 5,500 scenarios and 78,100 questions, each validated for realism and quality. Our comprehensive evaluation of ten state-of-the-art LLMs reveals that their average performance underperforms humans by 44.7\%, with performance degrading significantly when tracking and reasoning about the shift of mental states. This performance gap highlights fundamental limitations in current LLMs' ability to model the dynamic nature of human mental states.
View on arXiv@article{xiao2025_2505.17663,
title={ Towards Dynamic Theory of Mind: Evaluating LLM Adaptation to Temporal Evolution of Human States },
author={ Yang Xiao and Jiashuo Wang and Qiancheng Xu and Changhe Song and Chunpu Xu and Yi Cheng and Wenjie Li and Pengfei Liu },
journal={arXiv preprint arXiv:2505.17663},
year={ 2025 }
}