Lift3D-VLA: Lifting VLA Models to 3D Geometry and Dynamics-Aware Manipulation

Recently, Vision-Language-Action (VLA) models have demonstrated strong generalization across diverse tasks. However, effective robotic manipulation in physical environments fundamentally requires geometric understanding and spatial reasoning. While some VLA approaches attempt to incorporate 3D infor...