Scaling Parallel Sequence Models to Foundation-Scale Vision Encoders

• URL: http://arxiv.org/abs/2606.00746v1
• Date: Sat, 30 May 2026 14:29:43 GMT
• Title: Scaling Parallel Sequence Models to Foundation-Scale Vision Encoders
• Abstract summary: Vision foundation models are bottlenecked by the quadratic cost of self-attention.<n>We present C-GSPN based on 2D spatial propagation propagation.
• Score: 123.58723804218151
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Vision foundation models are bottlenecked by the quadratic cost of self-attention, which limits usable resolution and increases the cost of large-scale pretraining. Subquadratic alternatives such as linear attention and state-space models reduce this cost, but often serialize images into 1D token streams and weaken the 2D spatial structure important for vision. Generalized Spatial Propagation Networks (GSPN) instead propagate context directly on the 2D grid through line-scan recurrences, achieving near-linear complexity without positional embeddings, but have seen little use as foundation-scale encoders. We present C-GSPN, a foundation-scale vision encoder based on 2D spatial propagation. C-GSPN makes the operator practical through three improvements: (1) a fast GSPN CUDA kernel that fuses per-step launches into a single warp-specialized implementation with shared-memory tiling, coalesced access, and a compact multi-channel propagation, reaching over 90% of peak memory bandwidth and running up to 40--52x faster than the original GSPN implementation; (2) a compressed latent-space propagation block with fused normalization, which turns kernel-level speed into block- and model-level efficiency; and (3) a two-stage cross-operator distillation recipe that trains the new architecture from an attention teacher without the cost of from-scratch foundation-scale training. Distilled with 600M image-text pairs, C-GSPN matches an isomorphic ViT baseline with 15% fewer parameters, improves ADE20K segmentation by +2.1%, transfers to high resolution with a fraction of the data needed from scratch, and delivers a 4x end-to-end block speedup at 2K with single-pass, tiling-free inference.

Related papers

• OScaR: The Occam's Razor for Extreme KV Cache Quantization in LLMs and Beyond [50.440302567029654]
  Multi-modal intelligence has made the Key-Value cache a dominant memory bottleneck for efficient deployment.<n>In this work, we revisit the inherent limitations of the per-channel quantization paradigm.<n>We propose OScaR, an accurate and lightweight KV cache compression framework for X-LLMs.
  arXiv  Detail & Related papers  (2026-05-19T10:53:03Z)
• Neural Cluster First, Route Second: One-Shot Capacitated Vehicle Routing via Differentiable Optimal Transport [3.248584983235657]
  We introduce Neural CFRS, the first purely non-autoregressive one-shot neural CFRS framework for the Capacitated Vehicle Routing Problem (CVRP)<n>It enforces global fleet-capacity constraints end-to-end via a differentiable entropic Optimal Transport layer, producing a continuous transport plan to sparsify an exact capacitated solver.<n>By equipping the framework with a pre-trained vocabulary, we unlock extreme parameter efficiency and zero-shot scaling.
  arXiv  Detail & Related papers  (2026-05-10T03:53:59Z)
• BWTA: Accurate and Efficient Binarized Transformer by Algorithm-Hardware Co-design [71.97035034203275]
  We analyze zero-point distortion in binarization and propose a Binary Weights & Ternary Activations (BWTA) quantization scheme.<n>For training, we propose Smooth Multi-Stage Quantization, combining a Levelwise Degradation Strategy and a MagnitudeAlignment Projection Factor.<n>Experiments show that BWTA approaches full-precision performance for Transformer-based models, with an average 3.5% drop on GLUE and less than 2% drop on five tasks.
  arXiv  Detail & Related papers  (2026-04-05T04:25:07Z)
• Hierarchical Awareness Adapters with Hybrid Pyramid Feature Fusion for Dense Depth Prediction [9.96771578356719]
  Existing approaches rely on increasingly complex network architectures to regress depth maps.<n>We propose a multilevel conditional random field (CRF) model built upon the Swin Transformer backbone.<n>Our method achieves state-of-the-art performance, reducing Abs Rel to 0.088 ($-$7.4%) and RMSE to 0.316 ($-$5.4%) on NYU Depth v2, while attaining near-perfect threshold accuracy.
  arXiv  Detail & Related papers  (2026-04-03T07:59:26Z)
• GSPN-2: Efficient Parallel Sequence Modeling [101.33780567131716]
  Generalized Spatial Propagation Network (GSPN) addresses this by replacing quadratic self-attention with a line-scan propagation scheme.<n>GSPN-2 establishes a new efficiency frontier for modeling global spatial context in vision applications.
  arXiv  Detail & Related papers  (2025-11-28T07:26:45Z)
• Linear Differential Vision Transformer: Learning Visual Contrasts via Pairwise Differentials [34.77694214755808]
  We introduce Visual-Contrast Attention (VCA), a drop-in replacement for Multi-Head Self-Attention (MHSA)<n>VCA injects an explicit notion of discrimination while reducing the theoretical complexity from O(N N C) to O(N n C) with n N.<n>The module adds fewer than 0.3M parameters to a DeiT-Tiny backbone, requires no extra FLOPs, and is wholly architecture-agnostic.
  arXiv  Detail & Related papers  (2025-11-02T07:04:12Z)
• Parallel Sequence Modeling via Generalized Spatial Propagation Network [80.66202109995726]
  Generalized Spatial Propagation Network (GSPN) is a new attention mechanism for optimized vision tasks that inherently captures 2D spatial structures.<n>GSPN overcomes limitations by directly operating on spatially coherent image data and forming dense pairwise connections through a line-scan approach.<n>GSPN achieves superior spatial fidelity and state-of-the-art performance in vision tasks, including ImageNet classification, class-guided image generation, and text-to-image generation.
  arXiv  Detail & Related papers  (2025-01-21T18:56:19Z)
• Scale Attention for Learning Deep Face Representation: A Study Against Visual Scale Variation [69.45176408639483]
  We reform the conv layer by resorting to the scale-space theory. We build a novel style named SCale AttentioN Conv Neural Network (textbfSCAN-CNN) As a single-shot scheme, the inference is more efficient than multi-shot fusion.
  arXiv  Detail & Related papers  (2022-09-19T06:35:04Z)

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.