Large language model (LLM)-based agents interact with external tools through protocols such as the Model Context Protocol (MCP), producing tool-call traces t…
机构:上智院
来源:arXiv 2610.41885 | AI4Papers 论文推荐平台