SkillZip：通过发现可复用结构实现无评估的自进化代理技能压缩
原标题：SkillZip: Evaluation-Free Skill Compression for Self-Evolving Agents by Discovering Reusable Structure
AI 摘要
Hugging Face 每日论文发布了一篇题为《SkillZip: Evaluation-Free Skill Compression for Self-Evolving Agents by Discovering Reusable Structure》的论文。该论文提出了一种名为 SkillZip 的无评估技能压缩方法，通过发现可复用的结构来压缩自进化代理的技能，无需评估回滚。该方法基于最小描述长度目标，支持一次性模式和持续 Zip-on-Write 模式，实验证明其在压缩性能、泛化性和成本开销方面具有优势。
正文节选
SkillZip: Evaluation-Free Skill Compression for Self-Evolving Agents by Discovering Reusable Structure Abstract SkillZip compresses self-evolving agent skills by finding a minimal faithful structural explanation that shares repeated rules and procedures while preserving rare exceptions, without requiring evaluation rollouts. Self-evolving agents accumulate reusable skills by appending successful procedures and failure fixes. Over time, the same requirement is often restated in several branches,