Can vision-language models turn 2D observations into 3D understanding that supports precise action? This requires inferring object geometry, anticipating the…
机构:厦大
来源:arXiv 2610.17665 | AI4Papers 论文推荐平台