Self-Adaptive and Robust 6G Network Architecture Integrating Native GPTs

The emergence of generative pre-trained transform-ers (GPTs) will thoroughly change the application of sixth generation mobile communications (6G) networks. Therefore, it is necessary to design new network architectures to support ubiq-uitous deployment and real-time applications of GPTs. Aiming to integrate GPTs and the 6G network, this paper investigates the typical application scenarios of 6G+GPTs and summarizes the requirements of network key performance indicators (KPIs). Then, to address the complex and dynamically changing commu-nication environment, a self-adaptive 6G network architecture is proposed based on autonomous learning and self-optimization. Additionally, a novel mechanism based on attack samples is studied to improve the security of applying GPTs in 6G networks. Finally, we demonstrate that the proposed network architecture and security mechanism can satisfy the KPIs and improve robustness effectively. Overall, this paper provides a theoretical basis for the support of native GPTs with a novel 6G network architecture.