Scholarlink

With the rapid increase in the number and speed of scientific publications, researchers face significant time pressure when conducting literature reviews. This paper presents an automatic literature review generation method leveraging large language models (LLMs) and multi-agent systems (MAS). By designing multiple agent roles, including reference parsing, analysis, generation, and integration agents-this method fully utilizes the natural language processing capabilities of LLMs and the collaborative strengths of MAS to produce high-quality literature reviews. In the NLPCC2024 evaluation task, our method excelled in multiple automatic evaluation metrics (such as SoftHeadingRecall and ROUGE) and manual evaluations, showcasing its great potential for practical applications.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Generation of Scientific Literature Surveys Based on Large Language Models (LLM) and Multi-Agent Systems (MAS)

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ruihua Qi,
  • Weilong Li,
  • Haobo Lyu

摘要

With the rapid increase in the number and speed of scientific publications, researchers face significant time pressure when conducting literature reviews. This paper presents an automatic literature review generation method leveraging large language models (LLMs) and multi-agent systems (MAS). By designing multiple agent roles, including reference parsing, analysis, generation, and integration agents-this method fully utilizes the natural language processing capabilities of LLMs and the collaborative strengths of MAS to produce high-quality literature reviews. In the NLPCC2024 evaluation task, our method excelled in multiple automatic evaluation metrics (such as SoftHeadingRecall and ROUGE) and manual evaluations, showcasing its great potential for practical applications.