One of the central concerns for the safety of advanced AI is scheming: when a model strategically complies under oversight but pursues its own goals when uno…
机构:DeepMind
来源:arXiv 2610.26551 | AI4Papers 论文推荐平台