Proactive Detection of Breaking Dependency Updates using Large Language Models

A curated dataset saved in a jsonl file. Each data entry consists of: Package metadata API Diffs between two consecutive versions of the package Code diffs excerpt This dataset is used for evaluating the performance of LLMs in the task of classifying whether an update of a package is safe or breaking.本数据集为经人工精心整理甄选的数据集，存储于jsonl（JSON Lines）格式文件中。每条数据条目包含以下内容： 包元数据 软件包相邻两个版本间的API差异 代码差异节选 本数据集可用于评估大语言模型（Large Language Model，LLM）在分类软件包更新属于安全更新抑或破坏性更新这一任务中的性能表现。