Rank-Aware Hyperbolic Alignment for Vision-Language Dataset Distillation

• URL: http://arxiv.org/abs/2606.29464v1
• Date: Sun, 28 Jun 2026 15:41:31 GMT
• Title: Rank-Aware Hyperbolic Alignment for Vision-Language Dataset Distillation
• Abstract summary: Vision-language dataset distillation (VLDD) compresses a large image-text paired dataset into a small set of synthetic pairs.<n>We propose a rank-aware hyperbolic alignment (RAHA) that combines hierarchical geometry with explicit alignment-capacity control.<n>RAHA demonstrates competitive cross-modal retrieval and improved transfer indicators under fixed budgets.
• Score: 54.74045834035952
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Vision-language dataset distillation (VLDD) compresses a large image-text paired dataset into a small set of synthetic pairs that can efficiently train contrastive vision-language models under strict data and compute budgets. Most existing methods match expert trajectories or cross-modal statistics, yet still enforce full-dimensional alignment in a Euclidean embedding space. This is often overly restrictive due to rank-deficient image--text correlation, with shared semantics concentrated in a low-dimensional range and remaining variation spread across a weakly correlated residual subspace. LoRS relaxes alignment at the similarity level by low-rank factorization, but does not explicitly control dominant alignment capacity and structure in the representation space. We thus propose a rank-aware hyperbolic alignment (RAHA) that combines hierarchical geometry with explicit alignment-capacity control. RAHA lifts multimodal representations to hyperbolic space and optimizes distilled pairs with asymmetric objectives that enforce geodesic alignment in the shared range while regularizing the residual subspace to preserve modality-private diversity and improve transfer robustness. Experiments on benchmarks show that RAHA demonstrates competitive cross-modal retrieval and improved transfer indicators under fixed budgets.

Related papers

• GRASP: Geometry-aware Residual Alignment for Scalable Pretraining Data Attribution [5.770893169582546]
  This work reframes attribution as subset-level counterfactual utility prediction.<n>We introduce GRASP, an interaction-aware surrogate.<n>It decisively outperforms existing scalable baselines.
  arXiv  Detail & Related papers  (2026-06-05T04:17:50Z)
• Multimodal Distribution Matching for Vision-Language Dataset Distillation [50.411341509805936]
  Multimodal Distribution Matching is a geometry-aware framework for efficient and generalizable multimodal distillation.<n>MDM integrates complementary components at the data, model, and loss levels.<n>It yields compact synthetic sets that preserve multimodal semantics, substantially reduce distillation cost, and remain robust across architectures.
  arXiv  Detail & Related papers  (2026-05-22T10:41:58Z)
• Hyperbolic Enhanced Representation Learning for Incomplete Multi-view Clustering [57.38215918201251]
  We propose HERL, a Hyperbolic Enhanced Representation Learning framework for Incomplete Multi-View Clustering.<n> operating within the Poincaré ball, HERL constructs a structure-aware latent space to enhance representation learning.<n>We show that HERL consistently outperforms state-of-the-art approaches.
  arXiv  Detail & Related papers  (2026-04-18T10:50:46Z)
• INST-Align: Implicit Neural Alignment for Spatial Transcriptomics via Canonical Expression Fields [1.0487944945684993]
  INST-Align couples a coordinate-based deformation network with a shared Canonical Expression Field.<n>It achieves state-of-the-art mean OT Accuracy (0.702), NN Accuracy (0.719), and Chamfer distance, with Chamfer reductions of up to 94.9% on large-deformation sections.<n>It also yields biologically meaningful spatial embeddings and coherent 3D tissue reconstruction.
  arXiv  Detail & Related papers  (2026-04-13T21:44:18Z)
• SOTAlign: Semi-Supervised Alignment of Unimodal Vision and Language Models via Optimal Transport [43.640561199880274]
  Platonic Representation Hypothesis posits that neural networks converge toward a shared statistical model of the world.<n>Recent work exploits this convergence by aligning frozen pretrained vision and language models with lightweight alignment layers.<n>We ask whether meaningful alignment can be achieved with substantially less supervision.<n>We introduce a semi-supervised setting in which pretrained unimodal encoders are aligned using a small number of image-text pairs together with large amounts of unpaired data.
  arXiv  Detail & Related papers  (2026-02-26T18:55:06Z)
• Rethinking Multi-Condition DiTs: Eliminating Redundant Attention via Position-Alignment and Keyword-Scoping [61.459927600301654]
  Multi-condition control is bottlenecked by the conventional concatenate-and-attend'' strategy.<n>Our analysis reveals that much of this cross-modal interaction is spatially or semantically redundant.<n>We propose Position-aligned and Keyword-scoped Attention (PKA), a highly efficient framework designed to eliminate these redundancies.
  arXiv  Detail & Related papers  (2026-02-06T16:39:10Z)
• Modality Gap-Driven Subspace Alignment Training Paradigm For Multimodal Large Language Models [84.78794648147608]
  A persistent geometric anomaly, the Modality Gap, remains.<n>Prior approaches to bridge this gap are largely limited by oversimplified isotropic assumptions.<n>We propose the Fixed-frame Modality Gap Theory, which decomposes the modality gap into stable biases and anisotropic residuals.<n>We then introduce ReAlign, a training-free modality alignment strategy.
  arXiv  Detail & Related papers  (2026-02-02T13:59:39Z)
• Enhancing CLIP Robustness via Cross-Modality Alignment [54.01929554563447]
  We propose Cross-modality Alignment, an optimal transport-based framework for vision-language models.<n> COLA restores global image-text alignment and local structural consistency in the feature space.<n> COLA is training-free and compatible with existing fine-tuned models.
  arXiv  Detail & Related papers  (2025-10-28T03:47:44Z)

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.