Loading reusable skill documents into a bounded context window is now the primary way large language model (LLM) agents acquire task-specific capabilities, w…
机构:清华
来源:arXiv 2608.19993 | AI4Papers 论文推荐平台