aiqbench analysis
AVSplat addresses a critical bottleneck in 3D Gaussian Splatting by preventing performance degradation when processing dense multi-view inputs, a common issue where existing models suffer from attention dilution and over-smoothing. By introducing assist-view preconditioning and adaptive voxel fusion, the framework successfully scales view count without sacrificing fidelity, offering a significant step forward for efficient novel view synthesis in uncalibrated scenarios.
AVSplat: Dense-View Feed-Forward 3D Gaussian Splatting with Assist-View Preconditioning
arXiv:2609.05925v1 Announce Type: cross Abstract: Pose-free feed-forward 3D Gaussian Splatting enables novel view synthesis from uncalibrated multi-view images. Although more views should improve performance, existing methods often degrade with dense-view inputs because global aggregation spreads attention over many tokens, and naive voxel fusion averages many Gaussians into overly smooth representations. We present AVSplat, a framework that turns additional views into reliable signals for both aggregation and representation. Before global attention, each view performs a single lightweight interaction with a small set of Assist Views chosen for relevance and diversity, and the cached features provide a focused scene context that stabilizes correspondence. For representation, we use adaptive temperature-aware voxel fusion that sharpens attribution under high occupancy, guided by occupancy and point confidence. Crucially, AVSplat restores positive view scaling where performance remains stable or improves as more input views are added, instead of degrading in the dense-view regime. Ablations show that Assist View Preconditioning is primarily responsible for preventing dense-view degradation, while Occupancy-guided Voxel Fusion contributes most of the single-point image-quality gains.
Abstract:Pose-free feed-forward 3D Gaussian Splatting enables novel view synthesis from uncalibrated multi-view images. Although more views should improve performance, existing methods often degrade with dense-view inputs because global aggregation spreads attention over many tokens, and naive voxel fusion averages many Gaussians into overly smooth representations. We present AVSplat, a framework that turns additional views into reliable signals for both aggregation and representation. Before global attention, each view performs a single lightweight interaction with a small set of Assist Views chosen for relevance and diversity, and the cached features provide a focused scene context that stabilizes correspondence. For representation, we use adaptive temperature-aware voxel fusion that sharpens attribution under high occupancy, guided by occupancy and point confidence. Crucially, AVSplat restores positive view scaling where performance remains stable or improves as more input views are added, instead of degrading in the dense-view regime. Ablations show that Assist View Preconditioning is primarily responsible for preventing dense-view degradation, while Occupancy-guided Voxel Fusion contributes most of the single-point image-quality gains.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
arXiv:2609.05925 [cs.CV]
(or arXiv:2609.05925v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2609.05925 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.
Full text under the publisher's open licence · 265 words
Related coverage
Sparse auto-regressive modeling for scene generation from multi-view images
arXiv · Sep 4, 2026
UniqueSplat: View-conditioned 3D Gaussian Splatting for Generalizable 3D Reconstruction
arXiv · Aug 5, 2026
GrapeSplat: Geometry-Grounded Reconstruction via Amalgamated Pose-Free Encoding for Feed-Forward 3D Gaussian Splatting
arXiv · Sep 22, 2026
Per-View Gaussian Predictions Enable Training-Free Distractor Filtering in Feed-Forward 3DGS
arXiv · Aug 28, 2026