As Large Language Models (LLMs) evolve into autonomous agents that alter real-world states, ensuring operational safety across multi-step workflows has becom…
机构:港科大
来源:arXiv 2609.28197 | AI4Papers 论文推荐平台