RepoEval

该数据集名为RepoEval，是一个包含了最新且高质量的真实世界代码仓库的基准测试集，涵盖了行级、API调用以及函数体补全等多种场景。此外，RepoEval通过提供多种多样的场景，为代码补全模型的评估提供了便利。该数据集的任务是针对代码仓库级别的代码补全。

This dataset, named RepoEval, is a benchmark dataset containing up-to-date and high-quality real-world code repositories, covering various scenarios such as line-level completion, API call completion, and function body completion. Moreover, RepoEval facilitates the evaluation of code completion models by offering diverse scenarios. The core task of this dataset focuses on repository-level code completion.

搜集汇总

背景与挑战

背景概述

RepoEval是一个用于仓库级别代码补全任务的基准测试集，包含多种代码补全场景和真实世界的高质量仓库数据。它提供了评估代码补全性能的脚本和工具，支持迭代检索和生成方法。