让编程智能体自动学习并进化技能，提升代码任务解决能力。
CODESKILL: Learning Self-Evolving Skills for Coding Agents
- 用强化学习训练可自我演化的技能管理策略，从编码轨迹中提取多粒度技能。
- 在多个基准上平均通过率提升9.69%，技能库大小稳定不变。
- 适合需要持续优化的自动化编程系统或智能体研究者。
编程智能体在解决软件工程任务时会产生丰富的行为轨迹。为实现智能体的自我演化，这些轨迹可被提炼为可复用的程序化技能，以紧凑方式编码经验并指导未来行为。然而，现有技能构建与维护方法常依赖固定提示和启发式更新规则，难以明确知识如何选择、抽象与保留以最优服务下游智能体。我们提出 CODESKILL，一个基于大模型的框架，将技能提取与技能库维护重新建模为可学习的管理策略。CODESKILL 从编码智能体轨迹中提取多粒度程序化技能，随新经验演化技能，并维持一个紧凑的技能库以支持后续任务求解。我们使用强化学习训练 CODESKILL，采用混合奖励机制，结合密集的评分器式技能质量反馈与来自冻结下游智能体的稀疏可验证执行反馈。在 EnvBench、SWE-Bench Verified 与 Terminal-Bench 2 上的实验表明，CODESKILL 在平均通过率上比无技能基线提升 9.69%，比最强的基于提示或记忆的基线提升 4.01%，同时在迭代构建过程中保持技能库大小稳定。
原文摘要 · Abstract (English)
Coding agents produce rich trajectories while solving software-engineering tasks. To enable agent self-evolution, these trajectories can be distilled into reusable procedural skills that compactly encode experience to guide future behavior. However, existing skill construction and maintenance methods often rely on fixed prompts and heuristic update rules, leaving it unclear how knowledge should be selected, abstracted, and maintained to best serve downstream agents. We propose CODESKILL, an LLM-based framework that reformulates skill extraction and skill-bank maintenance as a learnable management policy. CODESKILL extracts multi-granularity procedural skills from coding-agent trajectories, evolves skills with new experience, and maintains a compact skill bank for future task solving. We train CODESKILL with reinforcement learning, using a hybrid reward that combines dense rubric-based skill-quality feedback with sparse verifiable execution feedback from the frozen downstream agent. Experiments on EnvBench, SWE-Bench Verified, and Terminal-Bench 2 show that CODESKILL improves average pass rate by 9.69 over the no-skill baseline and by 4.01 over the strongest prompt-based or memory baseline, while maintaining the skill bank at a stable size during iterative construction.
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。