TESTGENEVAL
收藏资源简介:
TESTGENEVAL是由卡内基梅隆大学和Meta AI的研究团队创建的一个大规模测试生成和测试完成基准数据集。该数据集包含68,647个测试,来自1,210个代码和测试文件对,涵盖11个维护良好的Python仓库。数据集的创建过程包括从SWEBench数据集中提取代码测试文件对,并通过覆盖率和变异分数进行评估。TESTGENEVAL主要用于评估模型在生成高质量测试套件和完成现有测试套件的能力,旨在解决软件测试中的代码覆盖和缺陷检测问题。
TESTGENEVAL is a large-scale test generation and test completion benchmark dataset created by research teams from Carnegie Mellon University and Meta AI. It contains 68,647 test cases sourced from 1,210 code-test file pairs, covering 11 well-maintained Python repositories. The dataset construction workflow involves extracting code-test file pairs from the SWEBench dataset, followed by evaluation using coverage and mutation scores. TESTGENEVAL is primarily intended to evaluate models' abilities to generate high-quality test suites and complete existing test suites, with the goal of addressing the challenges of code coverage and defect detection in software testing.




