This paper presents MobQA, a benchmark dataset designed to evaluate the semantic understanding capabilities of large language models (LLMs) for human mobility data through natural language question answering. While existing models predict human movement patterns well, their ability to interpret underlying reasons or semantic meaning remains unclear. To address this gap, we annotated 14,260 trajectory points from the GeoLife dataset with 30,703 semantic labels covering points of interest, mobility modes, and visit reasons, grounding semantics-dependent questions and reference answers. Building on these annotations, MobQA provides 5,800 question–answer pairs from daily and weekly GPS trajectories, spanning factual retrieval, multiple-choice semantic inference, and free-form interpretive explanation. Our evaluation of major LLMs reveals strong performance on factual retrieval but notable limits in semantic reasoning and explanation, with trajectory length substantially impacting effectiveness. These findings demonstrate both achievements and limitations of recent LLMs for semantic mobility understanding.

This paper presents Text2Traj2Text, a novel learning-by-synthesis framework for captioning possible contexts behind shopper’s trajectory data in retail stores. Our work will impact various retail applications that need better customer understanding, such as targeted advertising and inventory management. The key idea is leveraging large language models to synthesize a diverse and realistic collection of contextual captions as well as the corresponding movement trajectories on a store map. Despite learned from fully synthesized data, the captioning model can generalize well to trajectories/captions created by real human subjects. Our systematic evaluation confirmed the effectiveness of the proposed framework over competitive approaches in terms of ROUGE and BERT Score metrics.