Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Spatial-Aware VLA Pretraining through Visual-Physical Alignment from Human Videos

About

Vision-Language-Action (VLA) models provide a promising paradigm for robot learning by integrating visual perception with language-guided policy learning. However, most existing approaches rely on 2D visual inputs to perform actions in 3D physical environments, creating a significant gap between perception and action grounding. To bridge this gap, we propose a Spatial-Aware VLA Pretraining paradigm that performs explicit alignment between visual space and physical space during pretraining, enabling models to acquire 3D spatial understanding before robot policy learning. Starting from pretrained vision-language models, we leverage large-scale human demonstration videos to extract 3D visual and 3D action annotations, forming a new source of supervision that aligns 2D visual observations with 3D spatial reasoning. We instantiate this paradigm with VIPA-VLA, a dual-encoder architecture that incorporates a 3D visual encoder to augment semantic visual representations with 3D-aware features. When adapted to downstream robot tasks, VIPA-VLA achieves significantly improved grounding between 2D vision and 3D action, resulting in more robust and generalizable robotic policies.

Yicheng Feng, Wanpeng Zhang, Ye Wang, Hao Luo, Haoqi Yuan, Sipeng Zheng, Zongqing Lu• 2025

Related benchmarks

TaskDatasetResultRank
Robot ManipulationLIBERO
Goal Achievement97
494
Robotic ManipulationRoboCasa
Average Success Rate45.8
22
Put-Three-ObjReal Robot Tasks (Unseen environment)
Sub-task Success Rate44
4
Put-Three-Obj manipulationReal Robot Tasks Seen Environments
Sub-task Success Rate52
4
Water-Plant manipulationReal Robot Tasks Seen Environments
Sub-task Success Rate57
4
Wipe-BoardReal Robot Tasks (Unseen environment)
Sub-task Success Rate83
4
Wipe-Board manipulationReal Robot Tasks Seen Environments
Sub-task Success Rate0.83
4
Showing 7 of 7 rows

Other info

GitHub

Follow for update