EdgePrompt introduces a cloud-edge collaborative framework that enables efficient LLM deployment in 6G networks while enhancing security and reducing latency.

• Employs a hierarchical attention splicing mechanism that distributes computation between cloud and edge
• Implements privacy-preserving strategies that isolate sensitive information, reducing data leakage risks
• Achieves significant latency reduction while maintaining model performance
• Enables secure LLM integration into critical 6G infrastructure management

This research addresses core security challenges in deploying AI for network management by creating a framework that protects sensitive data while delivering the performance needed for real-time applications in 6G environments.

EdgePrompt: A Distributed Key-Value Inference Framework for LLMs in 6G Networks