Scholarlink

This study explores the potential of Large Language Models (LLMs) in generating automated test cases to enhance software testing and bug detection. By evaluating 100 GitHub issues, the research investigates whether LLM-generated tests could identify bugs prior to their resolution. Findings reveal that LLMs can be effective in certain cases, particularly with simpler code that has fewer dependencies and lower complexity metrics. However, the success of these tests diminishes as code complexity, imports, and function calls increase. The study highlights the potential of LLMs in software testing, while also noting the challenges associated with testing complex code structures.

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

Towards Automated Software Testing: Evaluating LLMs in Generating Effective Test Cases

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Marcel Pehlke,
  • Cindy Eudenbach,
  • Maximilian Graw,
  • Marc Jansen

摘要

This study explores the potential of Large Language Models (LLMs) in generating automated test cases to enhance software testing and bug detection. By evaluating 100 GitHub issues, the research investigates whether LLM-generated tests could identify bugs prior to their resolution. Findings reveal that LLMs can be effective in certain cases, particularly with simpler code that has fewer dependencies and lower complexity metrics. However, the success of these tests diminishes as code complexity, imports, and function calls increase. The study highlights the potential of LLMs in software testing, while also noting the challenges associated with testing complex code structures.