The field of radio localization and sensing is moving towards more accurate and efficient methods for non-line-of-sight (NLoS) environments and complex scenarios. Researchers are exploring the use of machine learning and data-driven approaches to improve localization accuracy and robustness. Notable developments include the use of conditional diffusion models for NLoS localization, LiDAR-driven machine learning frameworks for indoor surface classification, and ray-traced MIMO CSI datasets for high-precision radio map construction. These innovations have the potential to enable scalable and physically grounded solutions for various applications, including autonomous navigation, industrial automation, and emergency response. Noteworthy papers include: RadioDiff-Loc, which proposes a novel generative inference framework for NLoS localization based on conditional diffusion models. UrbanMIMOMap, which presents a large-scale urban MIMO CSI dataset generated using high-precision ray tracing. Ghost Points Matter, which develops a millimeter-wave radar system for far-range vehicle detection in tunnels, achieving high detection accuracy and real-time processing. BatStation, which presents a lightweight, in-situ radar sensing framework seamlessly integrated into 5G base stations, achieving robust performance across diverse radar types.