HuggingFace Daily Papers（社区热门论文）· HuggingFace Daily Papers（社区热门论文）·· 2026-07-03精选AI 评分72
SkillOpt-Lite：更快更好的智能体自我进化，只需一行代码
AI 导读
SkillOpt-Lite 将智能体技能优化形式化为零阶优化，提出文件系统轨迹探索、共识属性挖掘与独立验证门控三条原则。相比完整 SkillOpt，它加速收敛并在 LiveMath 上提升 GPT-5.5 达 +8.8 点、GPT-5.4-nano 达 +25.4 点，使 nano 模型超越标准 SkillOpt 优化的 GPT-5.4。该框架已集成至 VSCode Copilot，开发者仅需一行代码即可进化技能。框架还可泛化为完整工具链优化（HarnessOpt），在 SpreadsheetBench 上令 GPT-5.4-nano 达到 0.7758 准确率，超越运行标准流程的更大模型 GPT-5.5（0.7620）。代码已开源。
推荐理由
把复杂 agent 优化砍到一个最小可行管线，小模型靠它打大模型，VSCode Copilot 用一条 vibe 就能进化技能，做 agent 的人别错过。
正文
Abstract:While skill optimization for autonomous agents has gained traction, existing methods rely on complex pipelines. This leaves a fundamental question unaddressed: What constitutes a minimal viable pipeline for skill optimization, where every component is justified by theory or empirical necessity? We formalize skill optimization via Zeroth-Order (ZO) optimization, mapping classical counterparts (central difference, trust regions) to recent literature. Noting that unlike blind numerical perturbations in classical ZO, skill trajectories serve as interpretable debugging feedback. Grounded in Claude Code philosophy and PAC learning, we establish three principles for convergence and generalization: file-system-based trajectory exploration, consensus attribute mining, and independent validation gating. Eliminating redundancies, we propose SkillOpt-Lite. It accelerates convergence and outperforms full SkillOpt: improving LiveMath by +8.8 points on GPT-5.5 and +25.4 points on GPT-5.4-nano, allowing the nano model to surpass standard GPT-5.4 optimized by SkillOpt. Finally, we integrate our framework into production coding agents like VSCode Copilot, enabling developers to evolve agent skills via one line of vibe. Because our framework treats all agent components simply as standard editable code, this minimal pipeline naturally generalizes to full harness optimization (HarnessOpt). On SpreadsheetBench, HarnessOpt enables GPT-5.4-nano to achieve 0.7758 accuracy, outperforming the larger GPT-5.5 running standard pipelines (0.7620). Code is available at this https URL.
| Subjects: | Software Engineering (cs.SE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.03451 [cs.SE] |
| (or arXiv:2607.03451v1 [cs.SE] for this version) | |
| https://doi.org/10.48550/arXiv.2607.03451 arXiv-issued DOI via DataCite |
Submission history
From: Yifei Shen [view email]
[v1]
Fri, 3 Jul 2026 16:07:06 UTC (581 KB)
来源：HuggingFace Daily Papers（社区热门论文） · arxiv.org