Bridging Knowledge and Language Models in Healthcare: A RAG Survey

Abstract

In recent years, large-scale language models have emerged as a key driver of progress in healthcare-oriented artificial intelligence, substantially contributing to the advancement of modern AI systems in this domain. However, the limitations of purely generative models particularly in terms of accuracy and reliability have underscored the need for innovative solutions. Retrieval-Augmented Generation (RAG), which integrates information retrieval from external sources with text generation by language models, has substantially improved the accuracy, correctness, and trustworthiness of generated responses. This paper categorizes and reviews various types of RAG models, explores their applications in medicine, and introduces practical databases for implementing RAG in healthcare. It also outlines key criteria for evaluating the performance of RAG-based models. The aim of this review is to offer a comprehensive roadmap for researchers and developers to design and assess optimal, reliable RAG-driven solutions in healthcare through a deep understanding of RAG variants, use cases, relevant data sources, and evaluation metrics.

Ähnliche Arbeiten