PragmaticCode
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/microsoft/monitors4codegen
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了真实世界的开源Java项目，完整地包含了它们的开发环境和依赖项，旨在评估监控引导解码（MGD）。此外，PragmaticCode数据集包含了1924个方法和14234个解引用提示，特别设计以确保多样性和跨文件依赖性。该数据集的规模涵盖了151个代码仓库，任务专注于Java中的方法补全。
The PragmaticCode Dataset consists of real-world open-source Java projects, fully including their development environments and dependencies, and is designed to evaluate Monitor-Guided Decoding (MGD). Additionally, this dataset contains 1,924 methods and 14,234 dereference prompts, which are specifically constructed to ensure diversity and cross-file dependencies. Spanning 151 code repositories, the dataset focuses on the task of method completion in Java.
提供机构：
Curated by the authors
