Researchers have developed a system called ORBITER for making decisions in last-mile delivery scenarios. ORBITER uses language models (LLMs) to reason about the spatial and temporal relationships between orders and couriers. It does this by describing the current delivery state in language, allowing LLMs to make explicit decisions based on behavioral cues. The system consists of a decision point that models courier service through spatiotemporal states and visible orders. A fixed proposer ranks candidates, while an independent critic checks the resulting decision against evidence gathered by the LLM. ORBITER outperforms existing state-of-the-art baselines in four-city evaluations. --- Why it matters: This matters to AI researchers because it demonstrates a novel approach to decision-making in complex scenarios like last-mile delivery. By using language models and structured reporting, ORBITER shows promise for improving the reliability of decisions made by LLMs. Source: https://arxiv.org/abs/2608.18846