This book provides theoretical and practical knowledge of an LLM (Large Language Model)-based approach to metaheuristics. In this book, the basic theory and the latest techniques are explained in an easy-to-understand manner, with concrete examples. Another emphasis is its real-world applicability. The book presents empirical examples from practical data and show that the proposed approaches are successful when addressing tasks from the recent research areas such as (1) LLMs for EC (Evolutionary computation), (2) training LLMs for EC, (3) automated machine learning, and (4) program synthesis, etc., details of which will be provided in the appendix for the sake of readers study. These materials will include a description of available resources for readers interested in gaining hands-on experience with the subject. The fundamental themes of this book, therefore, include recent research on the promising combination of Generative AI, LLMs, evolutionary computation, and metaheuristics. The ultimate goal of this book is to enable readers to apply these ideas to artificial intelligence on their own. This book is intended for beginners interested in artificial intelligence and artificial life (from undergraduate to graduate students), researchers in related fields, and engineers considering their applications. Therefore, most topics in this book begin with accessible subjects that require no specialized knowledge, though some connect to unsolved problems and cutting-edge research themes.
Chapter 1 Introduction.
Chapter 2 Examples of using LLMs as
Metaheuristics.
Chapter 3 LLMs for Evolutionary Optimization.
Chapter 4
LLMs for Metaheuristics.
Chapter 5 Towards Scalable, Robust, and Open-ended
LLM-EC Integration.
Chapter 6 Conclusion.
Chapter 7 Appendix A: Basic
Tools.
Chapter 8 Appendix B: Case Study LLM for EC Operators.
Chapter 9
Appendix C: Case Study LLMs for AutoML.
Hitoshi Iba is a Professor at the Graduate School of Information Science and Technology at the University of Tokyo. From 1990 to 1998, he was a senior researcher at the Electro Technical Laboratory (ETL) in Ibaraki, Japan. He is a founding associate editor of the Journal of Genetic Programming and Evolvable Machines (GPEM) and was a founding associate editor of IEEE Transactions on Evolutionary Computation. He has published more than 100 papers and is a (co-)author of more than 20 books. He is also an underwater naturalist and experienced PADI divemaster, having completed about 1,300 dives.
João Eduardo Batista is a postdoctoral researcher at RIKEN-CCS, a leading research center in Japan for computational science and high-performance computing. He graduated with a PhD in Informatics from the Faculty of Sciences at the University of Lisbon in 2024, having researched the application of genetic programming for interpretable feature engineering in remote sensing. Currently, his research topics are attribution in LLMs and LLM optimization, as well as high-performance C code optimization using interpretable machine learning techniques.
Jinglue Xu is a researcher at Sakana AI, a Tokyo-based artificial intelligence company focused on generative AI and evolutionary computation. He received his Ph.D. in Information Science and Technology from the University of Tokyo in 2025. His research interests include large language models (LLMs), autonomous agents, evolutionary computation, and AutoML. He has conducted multiple research projects exploring the combination of evolutionary computation, LLMs, and AutoML. Currently, he works at Sakana AI on developing more efficient evolutionary computation methods and their applications to LLMs.
