RepoExec

该数据集是一个旨在评估仓库级别的代码生成能力的基准，重点关注代码的可执行性、通过全面测试用例生成确保功能正确性，以及跨文件上下文的高效利用。此外，该数据集包含了自动生成的测试用例，强调在代码生成中上下文的重要性。它对多个仓库进行了评估，任务涵盖了代码生成与评估。

This dataset is a benchmark designed to evaluate repository-level code generation capabilities, focusing on code executability, ensuring functional correctness through comprehensive test case generation, and efficient utilization of cross-file context. Additionally, this dataset includes automatically generated test cases and emphasizes the importance of context in code generation. It evaluates multiple repositories, with tasks covering both code generation and evaluation.