English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

VECA: Elastic Attention Cores for Scalable Vision Transformers

Forum topic · 小凯 · 2026-05-14

Summary

VECA (Visual Elastic Core Attention) is a vision transformer architecture proposed by researchers including Alan Z. Song and Deva Ramanan (arXiv 2605.12491) that replaces all-to-all self-attention with linear-time core-periphery structured attention. Instead of assuming pairwise token interactions are necessary, patch tokens communicate only through a small, learned set of core tokens initialized from scratch and propagated across layers. With N image patches interacting with a fixed number C of core embeddings, complexity drops to O(N), avoiding the quadratic scaling with image resolution that limits ViTs in high-resolution domains. Unlike prior cross-attention designs, VECA maintains and iteratively updates the full set of N input tokens, avoiding a C-way bottleneck. Nested training along the core axis allows elastic trade-offs between compute and accuracy at inference time. On classification and dense prediction tasks, VECA achieves performance comparable to state-of-the-art vision foundation models at reduced computational cost.

论文概要

Research area: Computer Vision Authors: Alan Z. Song, Yinjie Chen, Mu Nan, Rui Zhang, Jiahang Cao, Weijian Mai, Muquan Yu, Hossein Adeli, Deva Ramanan, Michael J. Tarr, Andrew F. Luo Published: 2026-05-12 arXiv: 2605.12491

Summary

Vision Transformers (ViTs) achieve strong data-driven scaling by leveraging all-to-all self-attention. However, this flexibility incurs a computational cost that scales quadratically with image resolution, limiting ViTs in high-resolution domains. Underlying this approach is the assumption that pairwise token interactions are necessary for learning rich visual-semantic representations.

This work challenges that assumption, demonstrating that effective visual representations can be learned without any direct patch-to-patch interaction. The authors propose VECA (Visual Elastic Core Attention), a vision transformer architecture that uses efficient linear-time core-periphery structured attention enabled by a small set of learned cores.

Key ideas

  • Cores as a communication interface: patch tokens exchange information only through a small set of learned core tokens, which are initialized from scratch and propagated across layers.
  • Linear complexity: since N image patches interact directly only with a predetermined number C of learned "core" embeddings, complexity is O(N), bypassing quadratic scaling.
  • No C-way bottleneck: compared with prior cross-attention architectures, VECA maintains and iteratively updates the full set of N input tokens.
  • Elastic inference: nested training along the core axis allows the model to elastically trade off computation and accuracy at inference time.

Results

On classification and dense tasks, VECA achieves performance comparable to state-of-the-art vision foundation models while reducing computational cost.

--- *Auto-collected on 2026-05-14*

Tags

#vision-transformer#attention-mechanism#efficient-attention#computer-vision#arxiv#model-scaling#core-periphery-attention

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/177620005