LLM agents learn by interacting with environments, yet these environments are hand-built and static: blind to an agent's weaknesses, and quickly left behind …
机构:Google
来源:arXiv 2608.19880 | AI4Papers 论文推荐平台