Hybrid LLM-Guided Search for Quantum Reservoir Architecture Design

This paper treats quantum reservoir computing (QRC) architecture design as a constrained black-box search problem and compares five policies under identical evaluation budgets: random search, evolutionary search, Bayesian/TPE optimization, a feedback-based LLM agent, and a hybrid method. The hybrid combines LLM proposals with memory, mutation, crossover, duplicate avoidance, and exploration. Across NARMA10, Mackey-Glass forecasting, and temporal parity, it ranks first on NARMA10 and temporal parity and second on Mackey-Glass. With 25 evaluations and three seeds, it improves over random search on every task, including a 23.6% relative Mackey-Glass error reduction.