analiza aiqbench
GS-VLA introduces a lightweight 3D Gaussian Splatting module that enables frozen Vision-Language-Action policies to adapt robustly to camera viewpoint shifts without retraining. By treating observation-space adaptation as a localized novel-view synthesis problem, this approach mitigates significant performance drops caused by minor physical displacements of robotic cameras. The method offers an efficient alternative to computationally expensive fine-tuning or generative data augmentation for improving deployment reliability in dynamic environments.
GS-VLA: Plug-and-Play Viewpoint Canonicalization for Frozen VLA Policies via Gaussian Splatting
arXiv:2608.19066v1 Announce Type: cross Abstract: This paper proposes a lightweight, plug-and-play framework that improves robustness to viewpoint shifts in Vision-Language-Action (VLA) policies without policy retraining. To our knowledge, this is the first approach to directly leverage 3D Gaussian-based novel-view synthesis for observation-space adaptation in VLA policies. Current VLA performance relies on the implicit assumption that training and deployment camera configurations are identical. Our experiments show that even a small displacement of the camera mount can reduce the success rate on the LIBERO benchmark from about 90% to about 10% in the worst case. Prior approaches, such as large-scale fine-tuning or generative data augmentation, are computationally expensive and risk catastrophic forgetting. To address this, viewpoint shifts are reformulated as a localized novel-view synthesis problem. Under a Locality assumption, that camera perturbations remain within a small bounded region relative to the workspace, viewpoint normalization reduces to a scene- and policy-independent disocclusion task. Our work implements this idea with a 4M-parameter 3D-Gaussian canonicalizer prepended to a frozen VLA policy. Without modifying policy weights, GS-VLA improves performance across three orthogonal axes: (1) Policy architectures, (2) Unseen task suites, and (3) Perturbation scales. These results show that a lightweight visual module can recover a large fraction of the performance lost under viewpoint shift, without policy retraining.
Abstract:This paper proposes a lightweight, plug-and-play framework that improves robustness to viewpoint shifts in Vision-Language-Action (VLA) policies without policy retraining. To our knowledge, this is the first approach to directly leverage 3D Gaussian-based novel-view synthesis for observation-space adaptation in VLA policies. Current VLA performance relies on the implicit assumption that training and deployment camera configurations are identical. Our experiments show that even a small displacement of the camera mount can reduce the success rate on the LIBERO benchmark from about 90% to about 10% in the worst case. Prior approaches, such as large-scale fine-tuning or generative data augmentation, are computationally expensive and risk catastrophic forgetting. To address this, viewpoint shifts are reformulated as a localized novel-view synthesis problem. Under a Locality assumption, that camera perturbations remain within a small bounded region relative to the workspace, viewpoint normalization reduces to a scene- and policy-independent disocclusion task. Our work implements this idea with a 4M-parameter 3D-Gaussian canonicalizer prepended to a frozen VLA policy. Without modifying policy weights, GS-VLA improves performance across three orthogonal axes: (1) Policy architectures, (2) Unseen task suites, and (3) Perturbation scales. These results show that a lightweight visual module can recover a large fraction of the performance lost under viewpoint shift, without policy retraining.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
arXiv:2608.19066 [cs.CV]
(or arXiv:2608.19066v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2608.19066 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 306 cuvinte
Subiecte conexe
VG3S: Visual Geometry Grounded Gaussian Splatting for Semantic Occupancy Prediction
arXiv · 23 iul. 2026
Rendering-Free Lookahead for Question-Guided Active Vision
arXiv · 9 oct. 2026
SplatSearch: Instance Image Goal Navigation for Mobile Robots using 3D Gaussian Splatting and Diffusion Models
arXiv · 21 iul. 2026
OctoSplat: Hybrid OctoMap-Gaussian Splatting for Active Semantic Mapping and Phenotyping with Horticultural Robots
arXiv · 17 aug. 2026