📄 中文摘要
SWE-CI 是一种新颖的方法，旨在评估智能体在持续集成（CI）环境中维护代码库的能力。该方法通过设计一系列基准任务，测试智能体在代码质量、错误修复和功能扩展等方面的表现。研究表明，智能体在处理复杂代码库时的表现与人类开发者存在显著差异，尤其是在理解代码逻辑和进行有效的错误诊断方面。此外，SWE-CI 还提供了针对不同类型代码库的评估标准，帮助开发者更好地理解智能体的能力和局限性。该研究为未来智能体在软件开发中的应用提供了重要的参考依据。
SWE-CI：通过持续集成评估智能体在维护代码库中的能力
📄 English Summary
SWE-CI: Evaluating Agent Capabilities in Maintaining Codebases via CI
SWE-CI introduces a novel approach to evaluate the capabilities of agents in maintaining codebases within a Continuous Integration (CI) environment. It designs a series of benchmark tasks to assess agent performance in areas such as code quality, bug fixing, and feature enhancement. The findings indicate significant differences between agents and human developers, particularly in understanding code logic and effective error diagnosis when handling complex codebases. Additionally, SWE-CI provides evaluation criteria tailored for different types of codebases, aiding developers in comprehending the strengths and limitations of agents. This research offers critical insights for the future application of agents in software development.