一篇新发表在 arXiv 上的研究探讨了代码整洁度如何影响 AI 编码代理的性能。研究人员发现，虽然代码质量不影响 Anthropic 的 Claude Code 等代理的成功率，但它显著影响了它们的运行效率。更整洁的代码减少了 token 使用量和文件重访次数，这表明传统的软件可维护性原则对于 AI 驱动的开发仍然至关重要。 AI
影响 表明传统的代码可维护性原则对于优化 AI 编码代理性能和降低计算成本至关重要。
排序理由 该集群包含一篇详细介绍 AI 编码代理研究的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
As autonomous coding agents see rapid adoption, their evaluation has primarily focused on task completion rates holding the target codebase fixed. This leaves a critical question unanswered: does the structural and stylistic quality, or ``cleanliness'' of the underlying code affe…
As autonomous coding agents see rapid adoption, their evaluation has primarily focused on task completion rates holding the target codebase fixed. This leaves a critical question unanswered: does the structural and stylistic quality, or ``cleanliness'' of the underlying code affe…