Anonymous-999/LibEvoBench

该数据集是一个多配置集合，专注于NumPy、SciPy和PyTorch库的API（应用程序编程接口）分析。它包括多个子数据集，如API兼容性矩阵（记录API在不同版本中的存在性）、API字典（包含API的元数据信息，如类别、文档URI、版本范围等）、基准测试任务（T1、T2、T3，涉及代码生成、API存在性检测和签名生成任务）以及签名兼容性矩阵（追踪API签名在不同版本中的变化）。数据集旨在支持API演化研究、版本兼容性检查、代码补全和生成模型的基准测试。特征字段包括API名称、版本标识、库类型、代码字符串、文档字符串等，覆盖了从1.7到2.7等多个版本范围。This dataset is a multi-configuration collection focused on API (Application Programming Interface) analysis for the NumPy, SciPy, and PyTorch libraries. It includes several subsets, such as API compatibility matrices (tracking the presence of APIs across different versions), API dictionaries (containing metadata like category, documentation URI, version ranges, etc.), benchmark tasks (T1, T2, T3, involving code generation, API existence detection, and signature generation tasks), and signature compatibility matrices (tracking changes in API signatures across versions). The dataset is designed to support research on API evolution, version compatibility checks, code completion, and benchmarking for generative models. Features include API names, version flags, library types, code strings, docstrings, and more, covering version ranges from 1.7 to 2.7.