aiqbench analysis
eVGGT addresses the critical latency bottleneck in real-world robotics by distilling geometry-aware visual encoders into a lightweight model. This approach delivers a fivefold speedup and significant memory reduction while enhancing success rates in imitation learning frameworks like ACT and Diffusion Policy. The work effectively bridges the gap between robust geometric priors and the inference speeds required for practical deployment.
eVGGT: An Efficient Geometry-Aware Vision Encoder for Visuomotor Policies
arXiv:2509.15880v2 Announce Type: replace Abstract: Geometry-grounded vision models, such as VGGT, have emerged as robust visual encoders, providing essential geometric priors for robotic manipulation. However, the high computational cost of these models often leads to slow inference, limiting their practical applications in real-world robotics. This paper introduces eVGGT, a lightweight geometry-aware vision encoder distilled from the high-performing VGGT. Our findings demonstrate two primary advantages: i) integrating eVGGT into imitation learning frameworks (including ACT and Diffusion Policy) yields up to a 6.3% improvement in success rate over standard 2D encoders across bimanual and single-arm tasks in both simulation and real-world settings with variable viewpoints; ii) eVGGT achieves a nearly 5 times speedup and a 63% reduction in memory usage compared to state-of-the-art geometry-aware encoders while maintaining comparable task performance. These results suggest that eVGGT substantially alleviates the performance-latency bottleneck that has limited geometry-aware visuomotor policies in real-world deployment.
Abstract:Geometry-grounded vision models, such as VGGT, have emerged as robust visual encoders, providing essential geometric priors for robotic manipulation. However, the high computational cost of these models often leads to slow inference, limiting their practical applications in real-world robotics. This paper introduces eVGGT, a lightweight geometry-aware vision encoder distilled from the high-performing VGGT. Our findings demonstrate two primary advantages: i) integrating eVGGT into imitation learning frameworks (including ACT and Diffusion Policy) yields up to a 6.3% improvement in success rate over standard 2D encoders across bimanual and single-arm tasks in both simulation and real-world settings with variable viewpoints; ii) eVGGT achieves a nearly 5 times speedup and a 63% reduction in memory usage compared to state-of-the-art geometry-aware encoders while maintaining comparable task performance. These results suggest that eVGGT substantially alleviates the performance-latency bottleneck that has limited geometry-aware visuomotor policies in real-world deployment.
IROS 2026. Project page: this https URL
Robotics (cs.RO)
arXiv:2509.15880 [cs.RO]
(or arXiv:2509.15880v2 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2509.15880 Focus to learn more arXiv-issued DOI via DataCite
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.
Full text under the publisher's open licence · 240 words
Related coverage
WNM-3D: A World Navigation Model with 3D Scene Conditioning for Closed-Loop VLN
arXiv · Aug 21, 2026
GeoScaffold: Learning Compact Geometric Latents via Reconstruction for Efficient Vision-Language Navigation
arXiv · Oct 5, 2026
BIND: Binding 3D Robot Actions to 2D Image Features
arXiv · Oct 1, 2026
RayViT: Ray-Conditioned Visual Representations for Viewpoint-Robust Imitation Learning
arXiv · Aug 3, 2026