分析AI编程助手如何改进代码性能，发现其修改模式高度集中。
What Do AI Agents Actually Change? An Empirical Taxonomy of Mutation Patterns in Performance-Improving Pull Requests
- 用双大模型交叉验证方法，对1254个性能相关代码变更分类
- 命名修改、对象创建、类型变更占主导，三类合计超86%
- 不同AI工具和优化目标使用不同的修改模式，可指导代码优化策略
AI编程代理是黑箱：我们无法观察其生成代码的内部过程，但可以分析其实际修改的内容。这对基于搜索的软件工程（SBSE）至关重要，因为遗传改进等技术依赖于反映真实代码变化的变异算子。在AIDev-pop中33,596个代理提交的拉取请求中，少于1%涉及性能优化，每个案例都是窥探代理行为的稀有窗口。我们对216个此类请求中的1,254个性能相关代码变更块进行了分类，采用Even-Mendoza等人（2025）提出的18类语法变异分类体系，通过双大模型交集管道完成。结果显示，名称修改（37.0%）、对象创建（26.4%）和类型变更（22.7%）三类主导，与以往遗传改进语料库中84%无变化的情况显著不同。每种代理系统有其独特的变异词汇表，每种性能优化策略激活的类别也基本不重叠。因此，代理身份和目标策略是缩小有效SBSE算子空间的重要先验信息。复现包：https://github.com/5uper6rain/ssbse-challenge-2026
原文摘要 · Abstract (English)
AI coding agents are black boxes: we cannot inspect how they generate code, but we can inspect what they change. This distinction matters for search-based software engineering (SBSE), where techniques such as genetic improvement (in the performance-optimisation application we study) depend on mutation operators that reflect how code is actually transformed. Fewer than 1% of the 33,596 agent PRs in AIDev-pop target performance, making each case a rare window into otherwise opaque agent behaviour. We classify 1,254 performance-relevant diff hunks from 216 of these PRs, spanning five agent systems, against the 18-category syntactic mutation taxonomy of Even-Mendoza et al. (2025) using a dual-LLM intersection pipeline. Three categories dominate: name modification (37.0%), object creation (26.4%), and type change (22.7%), a profile markedly different from prior GI corpora where no change accounted for 84%. Each agent's deployed system commits to a distinctive mutation vocabulary, and each performance strategy activates a largely disjoint category subset. Agent identity and target strategy are therefore informative priors that narrow the effective SBSE operator space. Replication package: https://github.com/5uper6rain/ssbse-challenge-2026
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。