Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement· 多日自主软件开发的持续改进框架
This paper studies autonomous software development, in which LLM-based coding agents transform high-level requirements into complete, functional, and usable software systems without human intervention. We introduce Harness-of-Harness (HoH), a framework that enables coding agents to continually improve software during autonomous development. HoH operates on existing coding-agent harnesses, and organizes their executions into iterative planning-coding-testing loops. To sustain improvement across loops, HoH balances repair with capability growth, scopes development into small and verifiable increments, separates implementation-time testing from independent evaluation, and constrains verifiable outputs rather than prescribing agent workflows. It progressively exposes deliverables, role-specifi
多日自主软件开发中持续改进的框架
- 核心方法
- 引入Harness-of-Harness (HoH) 框架,通过迭代规划-编码-测试循环持续改进软件,平衡修复和能力增长,分阶段进行可验证的小幅增量开发,分离实现时测试与独立评估,并约束可验证输出而非规定代理工作流程
- 适合谁读
- 研究者、工程师
- 要解决的问题
- 现有基于LLM的编码代理在自主软件开发中缺乏持续改进机制
- 关键实验
- 未提供
- 主要贡献
- 提供了一种在多日自主软件开发过程中持续改进软件质量的方法,提高了自主开发的效率和可靠性
- 意义与局限
- 意义在于推动了自主软件开发技术的进步,使其更加适用于复杂和长时间的项目开发;局限性在于目前可能仍然难以处理某些高度复杂的软件需求