Large Language Model Agents - (Machine Learning: Foundations, Methodologies, and Applications) (Hardcover)

Book Synopsis

What if you could learn large-model agents not by theory alone, but by building real, working systems from day one? This book offers exactly that: a practice-driven, code-grounded pathway into the rapidly evolving world of intelligent agents powered by large language models (LLM agents).

Designed for AI researchers, practitioners, and postgraduate students, the book reveals why large-model agents matter now--and how their architectures, memory mechanisms, retrieval-augmented generation (RAG), tool-use strategies, and reasoning enhancements come together to form deployable intelligent systems. Rather than repeating definitions, it highlights the real challenges professionals face: How do you design an agent workflow that remains robust under noisy inputs? How do you debug a failing reasoning chain? How can an agent migrate smoothly across models or platforms? Rather than prescribing a single "best" solution, the book deliberately preserves open design questions and comparative experiments, encouraging readers to develop their own engineering judgment.

Across four parts--agent foundations, agent architecture, agent fine-tuning, and frontier topics such as multimodal agents, multi-agent systems, agent safety, and agent protocols--this book blends conceptual clarity with executable Python examples and reproducible notebooks. Readers will learn prompt engineering techniques, memory and retrieval strategies, instruction tuning, LoRA and quantization workflows, reinforcement fine-tuning, and practical evaluation and debugging methods.

With only basic Python knowledge, readers can follow the "examples first, principles underneath" approach to master the essential skills for building, evaluating, and optimizing LLM agents. By the end, they will possess a complete methodological toolkit for stepping confidently into cutting-edge agent research and real-world applications.

From the Back Cover

What if you could learn large-model agents not by theory alone, but by building real, working systems from day one? This book offers exactly that: a practice-driven, code-grounded pathway into the rapidly evolving world of intelligent agents powered by large language models (LLM agents).

Designed for AI researchers, practitioners, and postgraduate students, the book reveals why large-model agents matter now--and how their architectures, memory mechanisms, retrieval-augmented generation (RAG), tool-use strategies, and reasoning enhancements come together to form deployable intelligent systems. Rather than repeating definitions, it highlights the real challenges professionals face: How do you design an agent workflow that remains robust under noisy inputs? How do you debug a failing reasoning chain? How can an agent migrate smoothly across models or platforms? Rather than prescribing a single "best" solution, the book deliberately preserves open design questions and comparative experiments, encouraging readers to develop their own engineering judgment.

Across four parts--agent foundations, agent architecture, agent fine-tuning, and frontier topics such as multimodal agents, multi-agent systems, agent safety, and agent protocols--this book blends conceptual clarity with executable Python examples and reproducible notebooks. Readers will learn prompt engineering techniques, memory and retrieval strategies, instruction tuning, LoRA and quantization workflows, reinforcement fine-tuning, and practical evaluation and debugging methods.

With only basic Python knowledge, readers can follow the "examples first, principles underneath" approach to master the essential skills for building, evaluating, and optimizing LLM agents. By the end, they will possess a complete methodological toolkit for stepping confidently into cutting-edge agent research and real-world applications.

About the Author

Munin Wen is an Research Assistant Professorat the School of Artificial Intelligence, Shanghai Jiao Tong University, and holds a PhD in Computer Science and Technology from the same institution. His research focuses on reinforcement learning, large-model agents, and multi-agent systems, with extensive hands-on experience in building and evaluating agent-based systems. He has published more than twenty papers in international venues and has served as a reviewer for NeurIPS, ICML, and ICLR since 2022.

Jianghao Lin is an Assistant Professor at the Antai College of Economics and Management, Shanghai Jiao Tong University. He received his PhD in Computer Science and Technology from Shanghai Jiao Tong University. His work centers on generative artificial intelligence and data science, with applications in recommender systems, operations optimization, and intelligent business analytics. He has authored over forty international publications, received two Best Paper Awards, and has one ESI Top 1% highly cited paper.

Weinan Zhang is a Professor and PhD advisor at Shanghai Jiao Tong University, where he teaches machine learning and reinforcement learning in the ACM Honors Class. He earned his PhD in Computer Science from University College London (UCL). His research spans reinforcement learning, data mining, knowledge graphs, and deep learning, with applications in recommender systems, game AI, and robotic control. He has published more than 180 papers in leading journals and conferences and has received the Wu Wenjun AI Outstanding Youth Award, the Alibaba DAMO Academy Young Fellow Award, and support from the China Association for Science and Technology's Young Talent Program.

Yong Yu is a Distinguished Professor at Shanghai Jiao Tong University, a recipient of the State Council Special Allowance, and one of the inaugural National High-Level Talent Program Teaching Masters. He is the founder of the ACM Honors Class and Director of the APEX Data and Knowledge Management Lab. His numerous honors include the National Model Teacher Award, the National Ethics in Teaching Award, the CCF Distinguished Education Award, the Shanghai May Day Labor Medal, and the SJTU President's Award. In 2018, he founded the Boyu Artificial Intelligence Institute, pioneering an innovative AI curriculum that has trained many outstanding algorithm engineers and researchers.