Code2Skill：从代码大规模合成可验证智能体技能
原标题：Paper page - Grounded Skill Synthesis from Code at Scale for Agentic Intelligence
AI 摘要
该论文提出 Code2Skill，一个全自动流水线，可从源代码中提取可验证的程序性技能，构建出包含 1,006,822 条记录的 CodeSkillBank。在覆盖九种模型设置和八个基准的 72 项评估中，使用检索到的技能后模型平均提升 11.7%，并在 57 个案例中优于基线。研究还显示，从 AI 生成代码合成的技能通过率达 93.50%，略高于人类编写代码的 93.00%。
正文节选
Grounded Skill Synthesis from Code at Scale for Agentic Intelligence Abstract Code2Skill automatically extracts verifiable procedural skills from source code to improve agent performance before interaction experience accumulates. Reusable skills give agents transferable procedural knowledge, making scalable acquisition essential for extending agents beyond prior experience. Existing methods face two limitations: trajectory-based synthesis requires interactions with specific environments, while d