nebius/SWE-bench-extra
收藏Hugging Face2025-05-28 更新2024-12-21 收录
下载链接：
https://hf-mirror.com/datasets/nebius/SWE-bench-extra
下载链接
链接失效反馈官方服务：
资源简介：
SWE-bench Extra数据集可用于训练或评估专门解决GitHub问题的代理系统。该数据集基于SWE-bench基准的构建方法，包含来自1,988个Python仓库的6,415个Issue-Pull Request对。数据收集过程包括Issue和Pull Request的收集、过滤以及基于执行的验证。数据集结构包括多个字段，如instance_id、patch、repo等，并添加了meta列以指示实例是否符合“lite”标准。
SWE-bench Extra is a dataset used to train or evaluate agentic systems specializing in resolving GitHub issues. It is built based on the methodology of the SWE-bench benchmark and includes 6,415 Issue-Pull Request pairs sourced from 1,988 Python repositories. The data collection process involves gathering issues and pull requests, filtering instances, and execution-based validation. The dataset statistics provide mean, 75th percentile, and maximum values for various attributes. The dataset structure details the type and description of each field.
提供机构：
nebius
