Learning to complete tasks in unfamiliar environments with unknown rules remains a key challenge for LLM agents. Current LLM agents often record their discov…
机构:UC Berkeley
来源:arXiv 2609.39140 | AI4Papers 论文推荐平台