The reviewed record of science sign in
Pith

arxiv: 2503.12355 · v1 · pith:IFWDKKLG · submitted 2025-03-16 · cs.CV · cs.LG

Atlas: Multi-Scale Attention Improves Long Context Image Modeling

Kumar Krishna Agrawal , Long Lian , Longchao Liu , Natalia Harguindeguy , Boyi Li , Alexander Bick , Maggie Chung , Trevor Darrell
show 1 more author
Adam Yala
This is my paper

Reviewed by Pithpith:IFWDKKLGopen to challenge →

classification cs.CV cs.LG
keywords atlasfasterimagemodelingmulti-scaleaccuracyachievesacross
0
0 comments X
read the original abstract

Efficiently modeling massive images is a long-standing challenge in machine learning. To this end, we introduce Multi-Scale Attention (MSA). MSA relies on two key ideas, (i) multi-scale representations (ii) bi-directional cross-scale communication. MSA creates O(log N) scales to represent the image across progressively coarser features and leverages cross-attention to propagate information across scales. We then introduce Atlas, a novel neural network architecture based on MSA. We demonstrate that Atlas significantly improves the compute-performance tradeoff of long-context image modeling in a high-resolution variant of ImageNet 100. At 1024px resolution, Atlas-B achieves 91.04% accuracy, comparable to ConvNext-B (91.92%) while being 4.3x faster. Atlas is 2.95x faster and 7.38% better than FasterViT, 2.25x faster and 4.96% better than LongViT. In comparisons against MambaVision-S, we find Atlas-S achieves 5%, 16% and 32% higher accuracy at 1024px, 2048px and 4096px respectively, while obtaining similar runtimes. Code for reproducing our experiments and pretrained models is available at https://github.com/yalalab/atlas.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. UCSF-PDGM-VQA: Visual Question Answering dataset for brain tumor MRI interpretation

    cs.CV 2026-05 unverdicted novelty 7.0

    Introduces the UCSF-PDGM-VQA benchmark for brain tumor MRI and finds that state-of-the-art VLMs exhibit modality collapse by suppressing visual features in favor of language priors.

  2. UCSF-PDGM-VQA: Visual Question Answering dataset for brain tumor MRI interpretation

    cs.CV 2026-05 unverdicted novelty 7.0

    Introduces the UCSF-PDGM-VQA dataset of 2387 QA pairs from 473 glioma MRI studies and demonstrates that state-of-the-art VLMs exhibit modality collapse on multi-sequence 3D medical images.

  3. Jolia: Concept-Level Vision-Language Alignment for 3D CT Contrastive Learning

    cs.CV 2026-06 unverdicted novelty 6.0

    ConQuer augments global CLIP alignment with independent per-concept contrastive losses on anatomical regions extracted from reports, producing Jolia which outperforms CLIP baselines on classification, report generatio...

  4. Enhancing Fine-Grained Spatial Grounding in 3D CT Report Generation via Discriminative Guidance

    cs.CV 2026-04 unverdicted novelty 6.0

    DCP-PD improves macro F1 scores on CT report generation benchmarks and introduces a hierarchical location-aware evaluation protocol that reveals ongoing challenges in pathology spatial grounding.