Wireless communication systems have undergone a remarkable evolution, transitioning from the era of 1G to 5G, which has shifted the focus from merely connecting people to connecting everything. To facilitate the connection and service of ubiquitous intelligent agents, 6G and future wireless communication systems are poised to usher in the Internet of Intelligence era. They will evolve into a comprehensive infrastructure that surpasses mere connectivity, deeply integrating artificial intelligence (AI) and sensing capabilities, and exhibiting high heterogeneity across multiple dimensions, including scenario coverage, system architecture, and functional implementation.

In terms of scenario coverage, 6G will encompass a full spectrum of environments, spanning air, space, land, and sea. From a system architecture perspective, it will explore new frequency bands and incorporate large-scale antenna arrays to support emerging services. Functionally, it will achieve machine synesthesia by deeply integrating radio frequency sensing with multimodal sensing.

To tackle the challenges posed by complex tasks and unforeseen scenarios, AI is recognized as a pivotal enabling technology for 6G. However, current wireless AI algorithms are predominantly task-specific models, which are constrained by limited scalability and generalization capabilities. Therefore, there is a pressing need for a unified design framework. One such framework is the wireless communication and sensing foundation model, WiFo-2, proposed by Cheng Xiang's team at Peking University. This model achieves generalized wireless representations across diverse scenarios, system configurations, and wireless communication and sensing tasks through extensive pre-training. It significantly enhances model performance while reducing management and deployment costs.