With increasing autonomy, Large Language Models (LLMs) may engage in deceptive behaviors, such as scheming to pursue misaligned goals or sandbagging to escap…
机构:清华
来源:arXiv 2610.27577 | AI4Papers 论文推荐平台