Many real-world tasks (e.g., office workflows, scientific experimentation) require LLM agents to interact repeatedly with their environments for context-depe…
机构:上交
来源:arXiv 2609.29773 | AI4Papers 论文推荐平台