Reinforcement learning with verifiable rewards trains on cheap checks, such as unit tests, that a policy can learn to game, while a stronger check is too exp…
机构:清华
来源:arXiv 2610.35716 | AI4Papers 论文推荐平台