Retrieval‐augmented Generation for Materials Large Language Models

View at publisher Summarise with VIRAT
Authors: Xue Jiang, Yanjing Su
Cited by: 0
References: 54

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This chapter explores the integration of retrieval-augmented generation (RAG) in large language models (LLMs) for materials research, focusing on the design principles of LLM-powered research assistants. The authors highlight the exponential growth of LLMs and their advanced language processing capabilities, which are invaluable in synthesizing vast amounts of data from diverse sources such as research papers and experimental records. The chapter delves into the architecture of LLM-powered research assistants, emphasizing their role in streamlining literature reviews, experimental design, and data analysis. Key design principles include semantic-aware literature retrieval, knowledge graph-enhanced summarization, multiobjective optimization for synthesis routes, and adaptive characterization technique selection. Additionally, the chapter discusses the importance of anomaly detection and outlier handling in data analysis. Through these principles, the authors demonstrate how conversational AI can drive innovation in materials science by addressing the complex and interdisciplinary challenges inherent in the field. The chapter also examines the evolution and limitations of RAG systems, focusing on three paradigms: Naive, Advanced, and Modular RAG.

Fields of Study & Subjects

Machine Learning in Materials Science Topic Modeling Artificial Intelligence Applications

Publication details

Journal / source
Wiley eBooks
Pages
387–418
Publication date
17 March 2026
Language
EN
Document type
Book chapter