AI and the World's Hard Problems · Theoretical computer science · arXiv · 2026-10-23
Large Language Models Meet Bayesian Optimization: Towards Self-evolving Wireless Digital Twins
Wireless digital twins rely on predefined learning or optimization structures that can become unsuitable as tasks and operating conditions change. The authors combine a large language model with Bayesian optimization, using BO as a stable backbone while the LLM decides whether to retain or adapt its components based on the current task and accumulated experience. In simulations of Gaussian process kernel discovery for a multi-user beam-selection problem, the LLM-discovered kernel improved final BO performance by 19.7% on training tasks over the strongest standard kernel baseline.