Scholarlink

With the growing demands on Question Answering (QA) systems for IT operations and maintenance, efficiently building such systems to support real-world products and deliver business value remains a critical challenge. This paper presents LLMA4ITOps, a lightweight multi-agent framework that leverages off-the-shelf Large Language Models (LLMs) to efficiently build QA systems for IT operations and maintenance. The framework is composed of three specialized agents: a Triage Agent for task routing, a Retrieval Agent for accessing domain-specific and up-to-date information, and a Generator Agent for producing final answers. Designed to support both knowledge acquisition and troubleshooting tasks, LLMA4ITOps eliminates the need for time-consuming data preparing, model training or fine-tuning. Experimental results on a cloud computing dataset demonstrate that the framework achieves superior performance compared to existing LLM-based solutions and offers strong potential for real-world enterprise 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

LLMA4ITOps: A Lightweight LLM-Based Multi-Agent Framework for IT Operations and Maintenance

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Zhuoxuan Jiang,
  • Tianyang Zhang,
  • Haotian Zhang,
  • Yinong Xun,
  • Yang Liu,
  • Dehua Feng,
  • Wen Si,
  • Shaohua Zhang

摘要

With the growing demands on Question Answering (QA) systems for IT operations and maintenance, efficiently building such systems to support real-world products and deliver business value remains a critical challenge. This paper presents LLMA4ITOps, a lightweight multi-agent framework that leverages off-the-shelf Large Language Models (LLMs) to efficiently build QA systems for IT operations and maintenance. The framework is composed of three specialized agents: a Triage Agent for task routing, a Retrieval Agent for accessing domain-specific and up-to-date information, and a Generator Agent for producing final answers. Designed to support both knowledge acquisition and troubleshooting tasks, LLMA4ITOps eliminates the need for time-consuming data preparing, model training or fine-tuning. Experimental results on a cloud computing dataset demonstrate that the framework achieves superior performance compared to existing LLM-based solutions and offers strong potential for real-world enterprise applications.