地理时空大模型：概念、范式与实践
大语言模型的快速发展为地理信息科学带来了新的机遇，但通用大模型并不天然具备处理坐标参考、拓扑关系和尺度效应等空间约束的能力。地理时空大模型作为在多源、多粒度和多模态地理时空大数据上预训练的大规模模型，从预训练中涌现出地理表征、地理理解、地理推理等核心能力，并可通过智能体机制将这些能力转化为自主执行空间分析的行动能力。本文梳理地理时空大模型的概念内涵与研究框架，沿"感知—认知—推理—行动"主线归纳四类研究范式：地理实体表征、多模态地理场景理解、时空语义驱动的地理推理以及大模型驱动的地理智能体。地理时空大模型不是通用大模型在地理领域的简单应用，而是将地理信息科学领域知识与人工智能方法深度融合形成的新型研究方向。未来应完善多源地理时空数据的组织与评测基准，将空间关系和时空规律融入模型训练与结果校验，并提高GIS任务执行过程及空间结果的可验证性，以增强模型在土地利用分析、城市时空预测和空间决策等任务中的可靠性。
LandAI: Explainable Parcel-Level Land-Use Classification with Multimodal Reasoning
We present the LandAI framework, which enables explainable land-use analysis through geometrically guided reasoning. By introducing GRPO-L1, a multimodal reinforcement learning algorithm, we encourage the model to generate coordinate-based visual references via bounding boxes before producing semantic conclusions. This structured reasoning process helps mitigate conflicts between visual signals and textual priors. In the land-use sample-size comparison, LandAI achieves state-of-the-art classification accuracy with a sample budget equivalent to 25% of the land-use training sample count reported for the comparison model, showing potential for data-scarce regions. Furthermore, LandAI can help identify human label noise in large-scale datasets. Our results show that geometrically guided reasoning can improve the transparency, data efficiency, and reliability of geospatial AI for land-use analysis.
Interpretable Vision-Language Reasoning for Residential Renewal
Assessing the urban renewal urgency enables planners to prioritize interventions and allocate resources effectively. Current visual-based evaluation methods rely primarily on pattern recognition and fail to provide explicit diagnostic reasoning about observed conditions. Vision-language models (VLMs) offer a promising approach by incorporating intrinsic semantic knowledge and enabling interpretable visual reasoning.
