Reinforcement learning for computer-use agents increasingly relies on proactive reward agents that interact with the post-execution GUI environment to verify…
机构:清华
来源:arXiv 2610.26469 | AI4Papers 论文推荐平台