meituan/SWE-Cycle
收藏资源简介:
SWE-Cycle是一个全面的基准测试数据集,用于评估AI编码代理在完整软件工程生命周期任务中的表现。它将软件开发过程分解为四个独立阶段(环境设置、代码实现、测试生成)和一个端到端任务(完整周期)。数据集包含489条记录,覆盖Python、Go、C、Ruby、JavaScript、Rust、TypeScript、Java和PHP等9种编程语言。数据主要来自SWE-bench Verified、SWE-bench Pro和Multilingual三个来源。每个实例包含唯一标识符、GitHub仓库信息、基础提交哈希、代码补丁、测试补丁、问题描述、提示文本等详细字段。数据集支持静态评估和动态评估两种方法,用于全面评估AI代理在不同软件开发阶段的能力。
SWE-Cycle is a comprehensive benchmark for evaluating AI coding agents on full software engineering lifecycle tasks. It decomposes the software development process into four isolated phases (Environment Setup, Code Implementation, Test Generation) and one integrated end-to-end task (Full Cycle). The dataset contains 489 records covering 9 programming languages including Python, Go, C, Ruby, JavaScript, Rust, TypeScript, Java, and PHP. The data primarily comes from three sources: SWE-bench Verified, SWE-bench Pro, and Multilingual. Each instance includes detailed fields such as unique identifier, GitHub repository information, base commit hash, code patch, test patch, problem statement, hints text, etc. The dataset supports both static evaluation and dynamic evaluation methods for comprehensive assessment of AI agent capabilities across different software development phases.




