LLM-Driven Agent for Efficient, Evidence-Grounded Mobility Prediction
arXiv 2606.05130·medium signal
Chen et al. build an LLM-driven agent for individual-level mobility prediction that emphasizes efficiency and evidence grounding, aiming to beat supervised sequence models used in urban simulation and transport planning. The grounding focus addresses LLM hallucination in spatiotemporal forecasting. A concrete agentic application outside the usual coding/chat domains.