Understanding Geometric Representations in Self-Supervised Vision Transformers via Subspace Intervention

• URL: http://arxiv.org/abs/2607.01987v1
• Date: Thu, 02 Jul 2026 10:18:02 GMT
• Title: Understanding Geometric Representations in Self-Supervised Vision Transformers via Subspace Intervention
• Abstract summary: We introduce a controlled subspace intervention framework to investigate how self-supervised Vision Transformers (ViTs) encode dense geometric information.<n>By connecting internal encoding mechanics with downstream performance, these findings provide a basis for effective feature selection and lightweight decoder design.
• Score: 7.887469434417159
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We introduce a controlled subspace intervention framework to investigate how self-supervised Vision Transformers (ViTs) encode dense geometric information. While linear probing is widely used to assess geometric representations, it treats features as a black box, failing to disentangle the underlying topology. To address this issue, we decompose the weights of converged linear probes to isolate the low-rank subspaces containing explicit geometric signals using Singular Value Decomposition (SVD). Our perspective yields three key insights: (1) Pre-training objectives determine how features are encoded. DINOv2 aligns spatial features for efficient linear extraction, while Masked Autoencoders (MAE) tend to disperse these signals, requiring a broader spatial context. (2) Explicit geometric representations are highly compressible, suggesting dense predictive heads could potentially be constrained to low-rank subspaces with minimal performance loss. (3) The layer-wise task affinity suggests that geometric precision peaks at intermediate layers before yielding to semantic abstraction in the final layers. By connecting internal encoding mechanics with downstream performance, these findings provide a basis for effective feature selection and lightweight decoder design. The source code is available at https://github.com/Zhou-Weichen/Geosubprobe.

Related papers

• Emergent 3D Instance Segmentation from Self-Supervised Point Transformers [15.864488266763976]
  Unsupervised 3D instance segmentation of outdoor LiDAR scans has traditionally relied on handcrafted geometric priors.<n>We investigate whether a frozen, self-supervised point transformer already contains the structural information required to isolate object instances.<n>We put these findings into our method TokenGraph3D, a training-free segmenter that groups points via connected components on a key-similarity graph.
  arXiv  Detail & Related papers  (2026-08-16T15:20:57Z)
• From Extrinsic to Intrinsic: Geodesic-Guided Representation Learning for 3D Geometric Data [56.41479248637621]
  We introduce a novel 3D representation learning paradigm, namely textbfPRISM, for textbfPre-training.<n>PRISM learns embeddings by textbfRecovering the textbfIntrinsic textbfSurface geodesic textbfMetric.<n> Experiments demonstrate that our approach shows satisfactory accuracy, robustness, and high efficiency in geodesic distance prediction.
  arXiv  Detail & Related papers  (2026-06-01T13:54:26Z)
• Topo-GS: Continuous Volumetric Embedding of High-Dimensional Data via Topological Gaussian Splatting [4.621191941382847]
  Topo-GS is a framework that repurposes 3D Gaussian Splatting (3DGS) to cast multidimensional projection as a meshless volumetric reconstruction process.<n>We show that Topo-GS successfully transforms discrete scatter plots into continuous volumetric representations.
  arXiv  Detail & Related papers  (2026-05-16T14:21:08Z)
• Learning 3D Representations for Spatial Intelligence from Unposed Multi-View Images [81.94999489820974]
  UniSplat is a feed-forward framework for learning 3D representations from unposed multi-view images.<n>We introduce a dual-masking strategy that strengthens geometry induction in the encoder.<n>Second, we develop a coarse-to-fine Gaussian splatting strategy that reduces appearance-semantics inconsistencies.<n>Third, we introduce a pose-conditioned recalibration mechanism that interrelates predicted 3D point and semantic maps into the image plane.
  arXiv  Detail & Related papers  (2026-04-12T10:36:18Z)
• SpatialFly: Geometry-Guided Representation Alignment for UAV Vision-and-Language Navigation in Urban Environments [49.966170814478915]
  We propose a geometry-guided spatial representation framework for UAV VLN.<n>Running on RGB observations without explicit 3D reconstruction, SpatialFly introduces a geometry-guided 2D representation alignment mechanism.<n> Experimental results show that SpatialFly consistently outperforms state-of-the-art UAV VLN baselines across both seen and unseen environments.
  arXiv  Detail & Related papers  (2026-03-22T03:56:58Z)
• IGGT: Instance-Grounded Geometry Transformer for Semantic 3D Reconstruction [82.53307702809606]
  Humans naturally perceive the geometric structure and semantic content of a 3D world as intertwined dimensions.<n>We propose InstanceGrounded Geometry Transformer (IGGT) to unify the knowledge for both spatial reconstruction and instance-level contextual understanding.
  arXiv  Detail & Related papers  (2025-10-26T14:57:44Z)
• Cross-Modal Geometric Hierarchy Fusion: An Implicit-Submap Driven Framework for Resilient 3D Place Recognition [9.411542547451193]
  We propose a novel framework that redefines 3D place recognition through density-agnostic geometric reasoning.<n>Specifically, we introduce an implicit 3D representation based on elastic points, which is immune to the interference of original scene point cloud density.<n>With the aid of these two types of information, we obtain descriptors that fuse geometric information from both bird's-eye view and 3D segment perspectives.
  arXiv  Detail & Related papers  (2025-06-17T07:04:07Z)
• Flattening-Net: Deep Regular 2D Representation for 3D Point Cloud Analysis [66.49788145564004]
  We present an unsupervised deep neural architecture called Flattening-Net to represent irregular 3D point clouds of arbitrary geometry and topology. Our methods perform favorably against the current state-of-the-art competitors.
  arXiv  Detail & Related papers  (2022-12-17T15:05:25Z)
• Self-supervised Geometric Perception [96.89966337518854]
  Self-supervised geometric perception is a framework to learn a feature descriptor for correspondence matching without any ground-truth geometric model labels. We show that SGP achieves state-of-the-art performance that is on-par or superior to the supervised oracles trained using ground-truth labels.
  arXiv  Detail & Related papers  (2021-03-04T15:34:43Z)
• Boundary-Aware Geometric Encoding for Semantic Segmentation of Point Clouds [45.270215729464056]
  Boundary information plays a significant role in 2D image segmentation, while usually being ignored in 3D point cloud segmentation. We propose a Boundary Prediction Module (BPM) to predict boundary points. Based on the predicted boundary, a boundary-aware Geometric. GEM is designed to encode geometric information and aggregate features with discrimination in a neighborhood.
  arXiv  Detail & Related papers  (2021-01-07T05:38:19Z)
• Scan-based Semantic Segmentation of LiDAR Point Clouds: An Experimental Study [2.6205925938720833]
  State of the art methods use deep neural networks to predict semantic classes for each point in a LiDAR scan. A powerful and efficient way to process LiDAR measurements is to use two-dimensional, image-like projections. We demonstrate various techniques to boost the performance and to improve runtime as well as memory constraints.
  arXiv  Detail & Related papers  (2020-04-06T11:08:12Z)
• Cylindrical Convolutional Networks for Joint Object Detection and Viewpoint Estimation [76.21696417873311]
  We introduce a learnable module, cylindrical convolutional networks (CCNs), that exploit cylindrical representation of a convolutional kernel defined in the 3D space. CCNs extract a view-specific feature through a view-specific convolutional kernel to predict object category scores at each viewpoint. Our experiments demonstrate the effectiveness of the cylindrical convolutional networks on joint object detection and viewpoint estimation.
  arXiv  Detail & Related papers  (2020-03-25T10:24:58Z)

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.