{"title":"基于知识库的大型语言模型的辐射防护智能助手。","authors":"Ankang Hu, Kaiwen Li, Zhen Wu, Hui Zhang, Rui Qiu, Junli Li","doi":"10.1007/s00411-025-01124-4","DOIUrl":null,"url":null,"abstract":"<p><p>Radiation protection is a critical pillar supporting the use of nuclear energy and nuclear technologies. The radiation protection system has been established with the accumulation of knowledge and experience. However, it is challenging for an individual or even a committee to master related knowledge and experience comprehensively and meticulously. An intelligent assistant that possesses extensive knowledge and experience in radiation protection is eagerly required. In this work, we propose an intelligent assistant in radiation protection based on a Large Language Model (LLM) with a knowledge base. The assistant can provide reliable answers with references from authoritative publications. The assistant was developed using open-source toolkits and open-source LLMs, and demonstrated satisfying answers to professional queries. Users can obtain reliable answers with references through the web-based user interface (UI). The assistant is designed for local deployment and utilizes private datasets, thereby addressing issues related to privacy and data security. The effectiveness of the assistant was evaluated by comparing it with LLM applications with web search. The results show that our method with a much smaller number of model parameters can deliver more precise and pertinent responses within the domain of radiation protection than web search-based systems. This work is a preliminary attempt to establish an intelligent assistant in the field of radiation protection, and it shows the potential for using LLM to increase efficiency in radiation protection-related tasks.</p>","PeriodicalId":21002,"journal":{"name":"Radiation and Environmental Biophysics","volume":" ","pages":""},"PeriodicalIF":1.5000,"publicationDate":"2025-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Intelligent assistant in radiation protection based on large language model with knowledge base.\",\"authors\":\"Ankang Hu, Kaiwen Li, Zhen Wu, Hui Zhang, Rui Qiu, Junli Li\",\"doi\":\"10.1007/s00411-025-01124-4\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><p>Radiation protection is a critical pillar supporting the use of nuclear energy and nuclear technologies. The radiation protection system has been established with the accumulation of knowledge and experience. However, it is challenging for an individual or even a committee to master related knowledge and experience comprehensively and meticulously. An intelligent assistant that possesses extensive knowledge and experience in radiation protection is eagerly required. In this work, we propose an intelligent assistant in radiation protection based on a Large Language Model (LLM) with a knowledge base. The assistant can provide reliable answers with references from authoritative publications. The assistant was developed using open-source toolkits and open-source LLMs, and demonstrated satisfying answers to professional queries. Users can obtain reliable answers with references through the web-based user interface (UI). The assistant is designed for local deployment and utilizes private datasets, thereby addressing issues related to privacy and data security. The effectiveness of the assistant was evaluated by comparing it with LLM applications with web search. The results show that our method with a much smaller number of model parameters can deliver more precise and pertinent responses within the domain of radiation protection than web search-based systems. This work is a preliminary attempt to establish an intelligent assistant in the field of radiation protection, and it shows the potential for using LLM to increase efficiency in radiation protection-related tasks.</p>\",\"PeriodicalId\":21002,\"journal\":{\"name\":\"Radiation and Environmental Biophysics\",\"volume\":\" \",\"pages\":\"\"},\"PeriodicalIF\":1.5000,\"publicationDate\":\"2025-07-02\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Radiation and Environmental Biophysics\",\"FirstCategoryId\":\"93\",\"ListUrlMain\":\"https://doi.org/10.1007/s00411-025-01124-4\",\"RegionNum\":4,\"RegionCategory\":\"环境科学与生态学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"BIOLOGY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Radiation and Environmental Biophysics","FirstCategoryId":"93","ListUrlMain":"https://doi.org/10.1007/s00411-025-01124-4","RegionNum":4,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"BIOLOGY","Score":null,"Total":0}
Intelligent assistant in radiation protection based on large language model with knowledge base.
Radiation protection is a critical pillar supporting the use of nuclear energy and nuclear technologies. The radiation protection system has been established with the accumulation of knowledge and experience. However, it is challenging for an individual or even a committee to master related knowledge and experience comprehensively and meticulously. An intelligent assistant that possesses extensive knowledge and experience in radiation protection is eagerly required. In this work, we propose an intelligent assistant in radiation protection based on a Large Language Model (LLM) with a knowledge base. The assistant can provide reliable answers with references from authoritative publications. The assistant was developed using open-source toolkits and open-source LLMs, and demonstrated satisfying answers to professional queries. Users can obtain reliable answers with references through the web-based user interface (UI). The assistant is designed for local deployment and utilizes private datasets, thereby addressing issues related to privacy and data security. The effectiveness of the assistant was evaluated by comparing it with LLM applications with web search. The results show that our method with a much smaller number of model parameters can deliver more precise and pertinent responses within the domain of radiation protection than web search-based systems. This work is a preliminary attempt to establish an intelligent assistant in the field of radiation protection, and it shows the potential for using LLM to increase efficiency in radiation protection-related tasks.
期刊介绍:
This journal is devoted to fundamental and applied issues in radiation research and biophysics. The topics may include:
Biophysics of ionizing radiation: radiation physics and chemistry, radiation dosimetry, radiobiology, radioecology, biophysical foundations of medical applications of radiation, and radiation protection.
Biological effects of radiation: experimental or theoretical work on molecular or cellular effects; relevance of biological effects for risk assessment; biological effects of medical applications of radiation; relevance of radiation for biosphere and in space; modelling of ecosystems; modelling of transport processes of substances in biotic systems.
Risk assessment: epidemiological studies of cancer and non-cancer effects; quantification of risk including exposures to radiation and confounding factors
Contributions to these topics may include theoretical-mathematical and experimental material, as well as description of new techniques relevant for the study of these issues. They can range from complex radiobiological phenomena to issues in health physics and environmental protection.