LANCE: Efficient Large AI Model Co-inference at 6G Edge Networks

Purpose and goal

The project will deliver unified frameworks for scalable, low-latency, privacy-preserving, and energy-efficient LAIM co-inference in 6G networks.

Expected effects and result

This project proposes a novel unified framework for efficient and trustworthy LAIM coinference in 6G edge networks, addressing critical gaps in theory, multi-dimensional resource management, and practical deployment. The results include publications, trained researchers, knowledge and testbed.

Planned approach and implementation

First, this project proposes a theoretical foundation for LAIM inference, with rigorous LAIM performance analysis. Then the project pioneers multi-dimensional resource management, introducing integrated optimization across communication and computation resources to balance LAIM inference quality, latency, and energy consumption. Finally, The developed testbed, datasets and benchmark suite will contribute as open research infrastructure for long-term community benefit.