Cross-Modal Redundancy and the Geometry of Vision-Language Embeddings

• URL: http://arxiv.org/abs/2602.06218v2
• Date: Mon, 09 Feb 2026 20:29:54 GMT
• Title: Cross-Modal Redundancy and the Geometry of Vision-Language Embeddings
• Abstract summary: Vision-language models (VLMs) align images and text with remarkable success, yet the geometry of their shared embedding space remains poorly understood.<n>We operationalize the Iso-Energy Assumption, which exploits cross-modal redundancy, with an Aligned Sparse Autoencoder (SAE)<n>We find that this inductive bias changes the SAE solution without harming reconstruction, giving us a representation that serves as a tool for geometric analysis.
• Score: 11.831077025246664
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Vision-language models (VLMs) align images and text with remarkable success, yet the geometry of their shared embedding space remains poorly understood. To probe this geometry, we begin from the Iso-Energy Assumption, which exploits cross-modal redundancy: a concept that is truly shared should exhibit the same average energy across modalities. We operationalize this assumption with an Aligned Sparse Autoencoder (SAE) that encourages energy consistency during training while preserving reconstruction. We find that this inductive bias changes the SAE solution without harming reconstruction, giving us a representation that serves as a tool for geometric analysis. Sanity checks on controlled data with known ground truth confirm that alignment improves when Iso-Energy holds and remains neutral when it does not. Applied to foundational VLMs, our framework reveals a clear structure with practical consequences: (i) sparse bimodal atoms carry the entire cross-modal alignment signal; (ii) unimodal atoms act as modality-specific biases and fully explain the modality gap; (iii) removing unimodal atoms collapses the gap without harming performance; (iv) restricting vector arithmetic to the bimodal subspace yields in-distribution edits and improved retrieval. These findings suggest that the right inductive bias can both preserve model fidelity and render the latent geometry interpretable and actionable.

Related papers

• The Art of Interrogation: Consistency Amplifies Factuality in Spatial Reasoning [65.67266333751569]
  Current Large Reasoning Models (LRMs) exhibit remarkable general capabilities but significantly underperform in spatial reasoning tasks.<n>We propose a self-supervised reinforcement learning framework that targets the internal reasoning process without requiring ground-truth annotations.<n>We show that this label-free consistency training approaches the accuracy of models trained with ground-truth supervision and achieves similar generalization across diverse tasks and data domains.
  arXiv  Detail & Related papers  (2026-06-10T10:50:06Z)
• Self-Evolving Spatial Reasoning in Vision Language Models via Geometric Logic Consistency [53.28076739140119]
  We propose a framework that enforces logical consistency in Vision-Language Models (VLMs) through geometric and linguistic duality operations.<n>SAGE is model-agnostic, data-efficient compared to prior GRPO methods, and can be applied as a lightweight post-training stage to any existing VLM.<n> Experiments on video and spatial reasoning benchmarks demonstrate consistent improvements over strong baselines.
  arXiv  Detail & Related papers  (2026-05-18T10:05:21Z)
• Anisotropic Modality Align [91.23979617826926]
  Training multimodal large language models has long been limited by the scarcity of high-quality paired multimodal data.<n>Recent studies show that the shared representation space of pretrained multimodal contrastive models can serve as a bridge, enabling models to perform multimodal training with unimodal data.<n>The core obstacle lies in the persistent Modality Gap in the shared space.
  arXiv  Detail & Related papers  (2026-05-08T14:53:24Z)
• Better with Less: Tackling Heterogeneous Multi-Modal Image Joint Pretraining via Conditioned and Degraded Masked Autoencoder [34.73963627819185]
  High-resolution (HR) joint optical and synthetic aperture radar (SAR) pretraining seeks modality synergy to mutually enhance single-source representations.<n>We propose CoDe-MAE, pioneering a textitbetter synergy with less alignment philosophy.<n>CoDe-MAE successfully prevents representation degradation and establishes new state-of-the-art performance across diverse single- and bi-modal downstream tasks.
  arXiv  Detail & Related papers  (2026-04-18T10:23:00Z)
• Reasoning on the Manifold: Bidirectional Consistency for Self-Verification in Diffusion Language Models [30.607422839762553]
  Bidirectional Manifold Consistency (BMC) is a training-free, unsupervised metric.<n>BMC serves as a robust discriminator of solution validity without ground truth answer.<n>Our results establish intrinsic geometric stability as a robust indicator of correctness for dLLMs.
  arXiv  Detail & Related papers  (2026-04-17T10:17:16Z)
• The Inductive Bias of Convolutional Neural Networks: Locality and Weight Sharing Reshape Implicit Regularization [57.37943479039033]
  We study how architectural inductive bias reshapes the implicit regularization induced by the edge-of-stability phenomenon in gradient descent.<n>We show that locality and weight sharing fundamentally change this picture.
  arXiv  Detail & Related papers  (2026-03-05T04:50:51Z)
• Rethinking Graph Generalization through the Lens of Sharpness-Aware Minimization [18.577399823354806]
  We focus on a prevalent yet under-explored phenomenon in graph generalization, Minimal Shift Flip (MSF)<n>MSF occurs when test samples that slightly deviate from the training distribution are abruptly misclassified.<n>We propose an energy-driven generative augmentation framework (E2A) that leverages energy-guided latent perturbations to generate pseudo-OOD samples.
  arXiv  Detail & Related papers  (2026-02-09T16:20:07Z)
• MirrorLA: Reflecting Feature Map for Vision Linear Attention [49.41670925034762]
  Linear attention significantly reduces the computational complexity of Transformers from quadratic to linear, yet it consistently lags behind softmax-based attention in performance.<n>We propose MirrorLA, a geometric framework that substitutes passive truncation with active reorientation.<n>MirrorLA achieves state-of-the-art performance across standard benchmarks, demonstrating that strictly linear efficiency can be achieved without compromising representational fidelity.
  arXiv  Detail & Related papers  (2026-02-04T09:14:09Z)
• Simulated Adoption: Decoupling Magnitude and Direction in LLM In-Context Conflict Resolution [3.0242762196828448]
  Large Language Models (LLMs) frequently prioritize conflicting in-context information over pre-existing parametric memory.<n>We show that models do not "unlearn" or suppress the magnitude of internal truths but rather employ a mechanism of geometric displacement.
  arXiv  Detail & Related papers  (2026-02-04T06:13:11Z)
• Adaptive Dual Uncertainty Optimization: Boosting Monocular 3D Object Detection under Test-Time Shifts [80.32933059529135]
  Test-Time Adaptation (TTA) methods have emerged to adapt to target distributions during inference.<n>We propose Dual Uncertainty Optimization (DUO), the first TTA framework designed to jointly minimize both uncertainties for robust M3OD.<n>In parallel, we design a semantic-aware normal field constraint that preserves geometric coherence in regions with clear semantic cues.
  arXiv  Detail & Related papers  (2025-08-28T07:09:21Z)
• Unfolded Laplacian Spectral Embedding: A Theoretically Grounded Approach to Dynamic Network Representation [1.4747234049753448]
  A common approach is to learn time-varying node embeddings, whose effectiveness depends on satisfying key stability properties.<n>We propose Unfolded Laplacian Spectral Embedding, a new method that extends the Unfolded Adjacency Spectral Embedding framework to normalized Laplacians.<n>As a bonus, we establish a new Cheeger-style inequality that connects the embeddings to the conductance of the underlying dynamic graphs.
  arXiv  Detail & Related papers  (2025-08-18T07:13:53Z)
• Escaping Plato's Cave: JAM for Aligning Independently Trained Vision and Language Models [30.07172193932125]
  We show that the Joint Autoencoder Modulator (JAM) induces alignment even across independently trained representations.<n>Our findings offer theoretical insight into the structure of shared semantics and practical guidance for transforming generalist unimodal foundations into specialist multimodal models.
  arXiv  Detail & Related papers  (2025-07-01T21:43:50Z)
• D-SCo: Dual-Stream Conditional Diffusion for Monocular Hand-Held Object Reconstruction [74.49121940466675]
  We introduce centroid-fixed dual-stream conditional diffusion for monocular hand-held object reconstruction. First, to avoid the object centroid from deviating, we utilize a novel hand-constrained centroid fixing paradigm. Second, we introduce a dual-stream denoiser to semantically and geometrically model hand-object interactions.
  arXiv  Detail & Related papers  (2023-11-23T20:14:50Z)
• Understanding and Constructing Latent Modality Structures in Multi-modal Representation Learning [53.68371566336254]
  We argue that the key to better performance lies in meaningful latent modality structures instead of perfect modality alignment. Specifically, we design 1) a deep feature separation loss for intra-modality regularization; 2) a Brownian-bridge loss for inter-modality regularization; and 3) a geometric consistency loss for both intra- and inter-modality regularization.
  arXiv  Detail & Related papers  (2023-03-10T14:38:49Z)
• Robust Self-Supervised LiDAR Odometry via Representative Structure Discovery and 3D Inherent Error Modeling [67.75095378830694]
  We develop a two-stage odometry estimation network, where we obtain the ego-motion by estimating a set of sub-region transformations. In this paper, we aim to alleviate the influence of unreliable structures in training, inference and mapping phases. Our two-frame odometry outperforms the previous state of the arts by 16%/12% in terms of translational/rotational errors.
  arXiv  Detail & Related papers  (2022-02-27T12:52:27Z)
• GELATO: Geometrically Enriched Latent Model for Offline Reinforcement Learning [54.291331971813364]
  offline reinforcement learning approaches can be divided into proximal and uncertainty-aware methods. In this work, we demonstrate the benefit of combining the two in a latent variational model. Our proposed metrics measure both the quality of out of distribution samples as well as the discrepancy of examples in the data.
  arXiv  Detail & Related papers  (2021-02-22T19:42:40Z)

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.