EVENT DOSSIER
On September 7, 2026, arXiv published a review article titled “Wireless Foundation Models: State-of-the-Art Technologies and Open Challenges”. This study systematically reviewed the latest advancements in foundation models in the wireless field, analyzed their core capabilities and performance bottlenecks in communication, sensing, and collaborative tasks, and deeply explored the current key challenges such as data scarcity, limited computing resources, privacy security, and lack of standardization. The aim was to provide guidance for subsequent research and engineering applications.
无线基础模型（WFMs）作为从大规模无线数据中学习可重用表示并适应下游任务的新兴方法，其文献因模态、预训练目标及架构差异而呈现碎片化。该综述系统分析了 WFMs 在物理层的应用，将现有文献归纳为信号识别与解调、信道表征学习、射频感知与定位、波束管理及频谱感知与监控五大物理层任务家族，并单独考察多任务物理层模型。分析涵盖预训练、适配及评估机制，重点关注下游任务多样性以及分布内、部分偏移和分布外迁移的区别。现有证据表明 WFMs 提供了可重用无线表示的日益增强的证明，但该证据在不同任务家族和评估设置中差异显著。数据、模态、架构、预训练目标、适配协议及分布偏移的差异导致难以确定驱动迁移与泛化的设计选择。结论指出了改进数据可用性、评估严谨性、泛化能力、高效适配及现实部署的开放方向，旨在为理解当前 WFMs 格局及开发…