How Can Large Language Models Make Network Configuration Human Friendly?
This is the central question explored in the research team's award-winning paper "NetConfEval: Can LLM Facilitate Network Configuration?" which earned the prestigious IRTF/IETF Applied Networking Research Prize.
Emerging Generative AI Models
The study investigates how emerging generative AI models can simplify complex network management by interpreting high-level human requirements, developing routing algorithms, and generating low-level network configurations. Changjie Wang, a doctoral student at KTH, led the work.
"In the long term, we expect to build AI-assisted tools that not only help automatically create software configure network systems but also verify the correctness and reliability of the proposed implementations. This will contribute to more reliable, efficient, and scalable network management in the future," says Marco Chiesa, associate professor at KTH and Wang's main supervisor.
Award-winning research
"The IRTF/IETF Applied Networking Research Prize recognizes only a few papers annually, selected from leading conferences such as SIGCOMM, NSDI, CCS, and IMC. We are honoured that our work has been included among these impactful contributions. This recognition highlights the relevance of applied research that connects theoretical insights with practical challenges in the networking domain," says Wang.
The work is part of a larger effort within a Vinnova-funded project led by Marco Chiesa, in collaboration with RedHat, RISE, and SAAB, that aims to fully integrate AI intelligence into software systems.
"Changjie's research aligns with recent breakthroughs in AI development, bringing us closer to fully automated networked systems," Chiesa concludes.
The work is a joint effort of Changjie Wang (KTH), Mariano Scazzariello (RISE), Alireza Farshin (Nvidia), Simone Ferlin (RedHat), Dejan Kostić (KTH/RISE, co-supervisor of Wang), and Marco Chiesa.
Read the award-winning paper: NetConfEval: Can LLMs Facilitate Network Configuration?
Published at the ACM International Conference on emerging Networking EXperiments and Technologies (CoNEXT), 2024:
Kontakt:
