Streaming video understanding requires models to repeatedly decide whether to respond or continue observing. Recently, Group Relative Policy Optimization (GR…
机构:复旦
来源:arXiv 2611.06728 | AI4Papers 论文推荐平台