arxivcs.ROcs.CV2026-07-07
Lift3D-VLA: Lifting VLA Models to 3D Geometry and Dynamics-Aware Manipulation
Jiaming Liu, Qingpo Wuwu, Nuowei Han, Hao Chen, Zhuoyang Liu, Fan Fei, et al.
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…