The Moving Eye: Enhancing VLA Spatial Generalization via Hybrid Dynamic Data Collection
For robotic manipulation researchers, this data-centric approach addresses the fundamental problem of shortcut learning in VLA models, offering a practical solution to improve spatial generalization across diverse architectures.
The authors propose a hybrid data collection strategy combining continuous camera motion with diverse static viewpoints to improve spatial generalization in Vision-Language-Action (VLA) models. Their method reduces spurious correlations and enables generalization to unseen camera poses and object configurations, benefiting all evaluated architectures (ACT, Diffusion, Pi0, Gr00t).
Vision-Language-Action (VLA) models have shown remarkable promise in generalized robotic manipulation. However, their spatial generalization remains fragile. We argue that simply increasing the number of viewpoints is insufficient. Models often fall into the trap of Shortcut Learning, latching onto spurious correlations (e.g., fixed relative poses between objects or between the camera and robot base) rather than learning true spatial relationships. In this work, we propose a data-centric solution to enhance VLA spatial generalization. We utilize a dual-arm setup where one arm performs manipulation while the other serves as a mobile environmental camera. We systematically evaluate three data distribution patterns: Fixed, Multi-Fixed, and Moving Views. Our findings reveal that a hybrid strategy, combining continuous camera motion with diverse static viewpoints, yields the best performance by substantially reducing spurious correlations while maintaining training stability. Our experiments demonstrate that this strategy mitigates spurious correlations, enabling VLAs to generalize to unseen camera poses and object configurations where simply adding more static viewpoints fails. Crucially, we reveal that the susceptibility to shortcut learning and the struggle with spatial generalization are universal characteristics shared across diverse architectures. Consequently, all evaluated models (ACT, Diffusion, and VLA models including Pi0 and Gr00t) benefit significantly from our mixed data strategy.