发表机构
Duke University（杜克大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
提出RuleEvolve，一个自我进化编码规则框架，通过LLM驱动的变异和评判迭代优化规则池，在多个框架和基准上超越手动工程和提示优化基线。
AI 中文摘要
AI编码代理的性能高度依赖于其底层的编码规则。然而，现有的编码规则通常是手工制作且固定的，这使得过程劳动密集且往往次优。在这项工作中，我们提出了RuleEvolve，一个用于编码规则的自我进化框架。RuleEvolve维护一个候选编码规则池，并迭代地改进它们。在每次迭代中，它使用一个由LLM驱动的变异器模块从现有候选规则中生成变体，然后使用一个评判器模块评估这些变体，并用表现最佳的变体更新规则池。在两个编码代理框架、四个骨干LLM和三个基准上的广泛评估表明，RuleEvolve在生成代码的功能正确性、代码长度和/或生成成本（例如，使用的令牌）方面优于手动工程和现有的提示优化基线。
英文摘要
The performance of AI coding agents is highly dependent on their underlying coding rules. However, existing coding rules are typically hand-crafted and fixed, making the process labor-intensive and often suboptimal. In this work, we propose RuleEvolve, a self-evolving framework for coding rules. RuleEvolve maintains a pool of candidate coding rules and iteratively improves them. In each iteration, it employs an LLM-powered mutator module to generate variants from existing candidates, and then uses a judge module to evaluate these variants and update the pool with the best-performing ones. Extensive evaluations across two coding-agent frameworks, four backbone LLMs, and three benchmarks demonstrate that RuleEvolve outperforms both manual engineering and existing prompt optimization baselines in terms of functional correctness of the generated code, code length, and/or generation cost (e.g., tokens used).
CommentsAccepted by NeurIPS 2026
