FAUN-Eval
收藏资源简介:
FAUN-Eval数据集由哈尔滨工业大学(深圳)和ByteDance的研究团队创建,旨在评估大型语言模型在解决GitHub问题中的细粒度能力。该数据集包含300个条目,涵盖代码问答、故障定位和代码编辑三个核心任务。数据集通过GitHub API从30个知名仓库中收集,确保了数据的多样性和真实性。创建过程中,通过交叉引用和关键词验证方法对每个条目进行精心编译和验证,确保数据质量。FAUN-Eval数据集主要应用于软件工程领域,旨在解决复杂代码库中的问题,提升软件质量和用户体验。
FAUN-Eval dataset was developed by research teams from Harbin Institute of Technology (Shenzhen) and ByteDance, aiming to evaluate the fine-grained capabilities of large language models (LLMs) when solving GitHub issues. This dataset contains 300 entries covering three core tasks: code question answering, fault localization, and code editing. It was collected from 30 well-known repositories via the GitHub API, ensuring the diversity and authenticity of the data. During the creation process, each entry was carefully compiled and validated through cross-referencing and keyword verification methods to guarantee data quality. The FAUN-Eval dataset is mainly applied in the field of software engineering, aiming to solve problems in complex code bases, improve software quality and user experience.




