Large Language Model-based Assistant Agents for Physical Layer Optical Networks

We present and demonstrate large language model (LLM)-based assistants that can simplify the management of physical layer optical networks. Our simulation shows GPT-4-based assistant can manage fiber input power over a four-span transmission link composed of mixed fiber types, by following natural language instructions for fiber input power constraints.