The paper "Towards Urban General Intelligence: A Review and Outlook of Urban Foundation Models" by Weijia Zhang, Jindong Han, Zhao Xu, Hang Ni, Hao Liu, and Hui Xiong provides a comprehensive overview of Urban Foundation Models (UFMs) and their potential in advancing urban intelligence. The authors define UFM as large-scale models pre-trained on extensive, diverse urban data, characterized by multi-source, multi-granularity, and multi-modal data characteristics. These models exhibit advanced capabilities in contextual understanding, problem-solving, and adaptability, making them suitable for various urban tasks such as traffic management, public safety, and urban planning. The paper highlights the challenges in building UFM, including data integration, spatio-temporal reasoning, and versatility across diverse urban domains. It proposes a data-centric taxonomy to categorize existing UFM-related works based on urban data modalities and types. Additionally, the authors introduce a framework for constructing UFM, designed to overcome identified challenges and enhance generalizability. The paper also explores the application landscape of UFM, detailing their potential impact in various urban contexts. It reviews existing studies on language-based, vision-based, trajectory-based, and time series-based models, as well as multimodal models. The review covers pre-training and adaptation techniques, including prompt engineering, model fine-tuning, and hybrid pre-training methods. Key contributions of the paper include: - A comprehensive and systematic review of UFM. - Introduction of the concept of UFM and specific challenges in their development. - A data-centric taxonomy for UFM research. - A novel framework for constructing UFM. - Detailed exploration of UFM applications in different urban domains. The paper concludes by discussing the future directions and potential of UFM in realizing Urban General Intelligence (UGI), which aims to transform cities into more livable, resilient, and adaptive spaces.The paper "Towards Urban General Intelligence: A Review and Outlook of Urban Foundation Models" by Weijia Zhang, Jindong Han, Zhao Xu, Hang Ni, Hao Liu, and Hui Xiong provides a comprehensive overview of Urban Foundation Models (UFMs) and their potential in advancing urban intelligence. The authors define UFM as large-scale models pre-trained on extensive, diverse urban data, characterized by multi-source, multi-granularity, and multi-modal data characteristics. These models exhibit advanced capabilities in contextual understanding, problem-solving, and adaptability, making them suitable for various urban tasks such as traffic management, public safety, and urban planning. The paper highlights the challenges in building UFM, including data integration, spatio-temporal reasoning, and versatility across diverse urban domains. It proposes a data-centric taxonomy to categorize existing UFM-related works based on urban data modalities and types. Additionally, the authors introduce a framework for constructing UFM, designed to overcome identified challenges and enhance generalizability. The paper also explores the application landscape of UFM, detailing their potential impact in various urban contexts. It reviews existing studies on language-based, vision-based, trajectory-based, and time series-based models, as well as multimodal models. The review covers pre-training and adaptation techniques, including prompt engineering, model fine-tuning, and hybrid pre-training methods. Key contributions of the paper include: - A comprehensive and systematic review of UFM. - Introduction of the concept of UFM and specific challenges in their development. - A data-centric taxonomy for UFM research. - A novel framework for constructing UFM. - Detailed exploration of UFM applications in different urban domains. The paper concludes by discussing the future directions and potential of UFM in realizing Urban General Intelligence (UGI), which aims to transform cities into more livable, resilient, and adaptive spaces.