Beyond Gaussian Bottlenecks: Topologically Aligned Encoding of Vision-Transformer Feature Spaces

• URL: http://arxiv.org/abs/2604.28122v1
• Date: Thu, 30 Apr 2026 17:12:31 GMT
• Title: Beyond Gaussian Bottlenecks: Topologically Aligned Encoding of Vision-Transformer Feature Spaces
• Abstract summary: We propose S$2$VAE, a geometry-first latent learning framework that focuses on compressing and representing the latent 3D state of a scene.<n>In depth estimation, camera pose recovery, and point cloud reconstruction, we show that geometry-aligned hyperspherical latents consistently outperform conventional Gaussian bottlenecks.
• Score: 18.19675060489249
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Modern visual world modeling systems increasingly rely on high-capacity architectures and large-scale data to produce plausible motion, yet they often fail to preserve underlying 3D geometry or physically consistent camera dynamics. A key limitation lies not only in model capacity, but in the latent representations used to encode geometric structure. We propose S$^2$VAE, a geometry-first latent learning framework that focuses on compressing and representing the latent 3D state of a scene, including camera motion, depth, and point-level structure, rather than modeling appearance alone. Building on representations from a Visual Geometry Grounded Transformer (VGGT), we introduce a novel type of variational autoencoder using a product of Power Spherical latent distributions, explicitly enforcing hyperspherical structure in the bottleneck to preserve directional and geometric semantics under strong compression. Across depth estimation, camera pose recovery, and point cloud reconstruction, we show that geometry-aligned hyperspherical latents consistently outperform conventional Gaussian bottlenecks, particularly in high-compression regimes. Our results highlight latent geometry as a first-class design choice for physically grounded visual and world models.

Related papers

• Ground4D: Consistency-Aware 4D Reconstruction from Monocular Video [69.94904173314505]
  We propose Ground4D, a geometry-grounded framework built on two stages.<n>First, we reconstruct multi-view-consistent 3D geometry and camera poses from monocular video.<n>Second, we conduct geometry-consistency-aware refinement via dynamic Gaussian Splatting.
  arXiv  Detail & Related papers  (2026-06-27T09:27:42Z)
• Towards Realistic and Consistent Orbital Video Generation via 3D Foundation Priors [61.34273238077091]
  We present a novel method for generating orbital videos from a single image of an object.<n>Our method achieves superior visual quality, shape realism and multi-view consistency compared to state-of-the-art methods.
  arXiv  Detail & Related papers  (2026-04-14T05:35:46Z)
• SpatialStack: Layered Geometry-Language Fusion for 3D VLM Spatial Reasoning [22.547972947051765]
  Large vision-language models (VLMs) still struggle with reliable 3D spatial reasoning, a core capability for embodied and physical AI systems.<n>We propose SpatialStack, a hierarchical fusion framework that progressively aligns vision, geometry, and language representations across the model hierarchy.<n>Building upon this framework, we develop VLM-SpatialStack, a model that achieves state-of-the-art performance on multiple 3D spatial reasoning benchmarks.
  arXiv  Detail & Related papers  (2026-03-28T22:49:40Z)
• VGGRPO: Towards World-Consistent Video Generation with 4D Latent Reward [14.297118875382871]
  VGGRPO is a latent geometry-guided framework for geometry-aware video post-training.<n>It stitches video diffusion latents to geometry foundation models, enabling direct decoding of scene geometry from the latent space.<n>It improves camera stability, geometry consistency, and overall quality while eliminating costly VAE decoding.
  arXiv  Detail & Related papers  (2026-03-27T16:57:51Z)
• Generation Models Know Space: Unleashing Implicit 3D Priors for Scene Understanding [50.098085774845195]
  We propose a paradigm shift by leveraging the implicit spatial prior within large-scale generation models.<n>We introduce GeneEGA-3D (Video Extracted Generative Awareness), a plug-and-play framework that repurposes a pre-trained video diffusion model as a Latent World Simulator.
  arXiv  Detail & Related papers  (2026-03-19T17:59:58Z)
• RAYNOVA: Scale-Temporal Autoregressive World Modeling in Ray Space [51.441415833480505]
  RAYNOVA is a multiview world model for driving scenarios that employs a dual-causal autoregressive framework.<n>It constructs an isotropic-temporal representation across views, frames, and scales based on relative Plcker-ray positional encoding.
  arXiv  Detail & Related papers  (2026-02-24T08:41:40Z)
• GeoVideo: Introducing Geometric Regularization into Video Generation Model [46.38507581500745]
  We introduce geometric regularization losses into video generation by augmenting latent diffusion models with per-frame depth prediction.<n>Our method bridges the gap between appearance generation and 3D structure modeling, leading to improved structural coherence-temporal shape, consistency, and physical plausibility.
  arXiv  Detail & Related papers  (2025-12-03T05:11:57Z)
• GeoWorld: Unlocking the Potential of Geometry Models to Facilitate High-Fidelity 3D Scene Generation [68.02988074681427]
  Previous works leveraging video models for image-to-3D scene generation tend to suffer from geometric distortions and blurry content.<n>In this paper, we renovate the pipeline of image-to-3D scene generation by unlocking the potential of geometry models.<n>Our GeoWorld can generate high-fidelity 3D scenes from a single image and a given camera trajectory, outperforming prior methods both qualitatively and quantitatively.
  arXiv  Detail & Related papers  (2025-11-28T13:55:45Z)
• WorldGrow: Generating Infinite 3D World [75.81531067447203]
  We tackle the challenge of generating the infinitely extendable 3D world -- large, continuous environments with coherent geometry and realistic appearance.<n>We propose WorldGrow, a hierarchical framework for unbounded 3D scene synthesis.<n>Our method features three core components: (1) a data curation pipeline that extracts high-quality scene blocks for training, making the 3D structured latent representations suitable for scene generation; (2) a 3D block inpainting mechanism that enables context-aware scene extension; and (3) a coarse-to-fine generation strategy that ensures both global layout plausibility and local geometric/textural fidelity.
  arXiv  Detail & Related papers  (2025-10-24T17:39:52Z)
• Epipolar Geometry Improves Video Generation Models [73.44978239787501]
  3D-consistent video generation could significantly impact numerous downstream applications in generation and reconstruction tasks.<n>We explore how epipolar geometry constraints improve modern video diffusion models.<n>By bridging data-driven deep learning with classical geometric computer vision, we present a practical method for generating spatially consistent videos.
  arXiv  Detail & Related papers  (2025-10-24T16:21:37Z)
• LIST: Learning Implicitly from Spatial Transformers for Single-View 3D Reconstruction [5.107705550575662]
  List is a novel neural architecture that leverages local and global image features to reconstruct geometric and topological structure of a 3D object from a single image. We show the superiority of our model in reconstructing 3D objects from both synthetic and real-world images against the state of the art.
  arXiv  Detail & Related papers  (2023-07-23T01:01:27Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.