Application and frontier exploration of retrieval-augmented generation technology in medical artificial intelligence
10.12173/j.issn.1005-0698.202503219
- VernacularTitle:检索增强生成技术在医学人工智能中的应用与前沿探索
- Author:
Zhe JIN
1
;
Jian ZOU
;
Xiao LI
;
Jiaxin LYU
;
Zhongxu HU
;
Da FENG
Author Information
1. 华中科技大学同济医学院药学院(武汉 430030)
- Publication Type:Journal Article
- Keywords:
Large language model;
Retrieval-augmented generation;
Hallucination problem;
Medical information retrieval
- From:
Chinese Journal of Pharmacoepidemiology
2025;34(8):962-971
- CountryChina
- Language:Chinese
-
Abstract:
With the rapid rise of large language models(LLM),the natural language generation capabilities of deep learning have demonstrated significant value in the medical field.However,the"closed nature"of model parameters makes them prone to generating"hallucinations",making it difficult to provide accurate answers to the latest knowledge,and the reasoning process lacks transparency and traceability.Retrieval-augmented generation(RAG)technology addresses these issues by actively connecting external information sources such as document databases and knowledge graphs during the generation process.This significantly reduces the dependence of LLM on outdated training data and introduces verifiable evidence and real-time knowledge updates into their responses.In the medical field,RAG technology effectively addresses the high-accuracy and traceability requirements of literature retrieval and clinical decision support.It is widely applied in areas such as drug discovery,pharmacovigilance,and the diagnosis and treatment of rare diseases.By integrating emerging technologies such as reinforcement learning,multimodal processing,and compliant privacy protection,RAG technology is evolving towards a more open and highly customizable direction,providing innovative intelligent solutions for medical information retrieval and decision-making support.
