CommitPack

该数据集用于评估原始代码与生成代码之间的功能正确性和相似度。在评估过程中，我们采用了专注于功能和语义准确性的CodeBERTScore。该数据集的任务是对代码生成进行评估。

This dataset is designed to evaluate the functional correctness and similarity between original code and generated code. During the evaluation process, CodeBERTScore which focuses on functional and semantic accuracy is employed. The core task of this dataset is to assess code generation.