Agents
Self-Evolving Coding Rules for AI Coding Agents
arXiv:2610.00650v1 Announce Type: cross Abstract: The performance of AI coding agents is highly dependent on their underlying coding rules. However, existing coding rules are typically hand-crafted an
arXiv:2610.00650v1 Announce Type: cross Abstract: 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).
Related
- Mendel Godel Machine: Recursive Self-Improving Coding Agents via Comparative Evolution
- Learning Globally Reusable Skills for Coding Agents
- Self-Improving AI Coding Agents Through Accumulated Behavioral Rules: A Closed-Loop Framework
Source: arXiv cs.AI | 2026-10-02