Skip to content
AI News HubLIVE
Original source2 min read

Where Global Context and Weight Sharing Work in Compact Vision Backbones

Summary

This paper investigates where global self-attention should be inserted and whether repeated blocks can share weights in compact vision backbones like EfficientNet-B0. It finds that zero-gated attention adapters at stages P3–P5 improve ImageNet-1k accuracy by 0.954 percentage points, while alternating two shared block templates in Stage 6 cuts parameters by 10.9% with only a 1.84-point drop.

SourcearXiv Computer VisionAuthor: Michael Schroeder
Where Global Context and Weight Sharing Work in Compact Vision Backbones
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 17 Nov 2026]

Title:Where Global Context and Weight Sharing Work in Compact Vision Backbones

View a PDF of the paper titled Where Global Context and Weight Sharing Work in Compact Vision Backbones, by Michael Schroeder

View PDF HTML (experimental)

Abstract:Compact convolutional backbones such as EfficientNet achieve strong parameter efficiency through local depthwise separable convolutions, yet two structural questions remain central to hybrid design: where does global self-attention improve recognition over pure convolution, and can repeated blocks share weights without losing representational capacity? We investigate both questions on ImageNet-1k (50,000-image validation) using controlled EfficientNet-B0 experiments. First, we show that wrapping a classification backbone in multi-scale feature pyramids (BiFPN) fails to outperform a plain MBConv trunk (69.99% - 71.81% vs. 72.86% plain B0) until paired with stage attention (73.15%). In contrast, keeping the MBConv trunk intact and inserting lightweight zero-gated residual self-attention adapters at stages P3 (28x28 windowed), P4 (14x14 global), and P5 (7x7 global) reaches 74.02% standalone and outperforms a bit-for-bit matched plain-B0 control by +0.954 percentage points (73.88% vs. 72.92%) across an 88-epoch stateless replay schedule, with learned attention amplitude rising monotonically with semantic depth (27.2% at P3 -> 45.4% at P4 -> 66.3% at P5). Second, we evaluate recurrent weight sharing across repeated MBConv blocks with pass-specific BatchNorm statistics (+1,536 parameters per pass). While post-hoc kernel averaging degrades accuracy to 17.03% and single-template stage tying (-27.6% parameters) plateaus at 66.10%, alternating two shared block templates ([B1, A, B, A]) in Stage 6 cuts total model parameters by 10.9% (582K weights) and recovers by +10.82 points in a single epoch to 70.92% (-1.84 points vs. unshared control).

Comments: 8 pages, 5 figures

Subjects:

Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)

Cite as: arXiv:2611.23446 [cs.CV]

(or arXiv:2611.23446v1 [cs.CV] for this version)

https://doi.org/10.48550/arXiv.2611.23446

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Michael Schroeder [view email] [v1] Tue, 17 Nov 2026 19:49:55 UTC (101 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Where Global Context and Weight Sharing Work in Compact Vision Backbones, by Michael Schroeder

View PDF

HTML (experimental)

TeX Source

view license

Additional Features

Audio Summary

Current browse context:

cs.CV

new | recent | 2026-11

Change to browse by:

cs cs.AI cs.LG

References & Citations

SciX

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

ResearchersAdvanced

Key points

  • Zero-gated self-attention adapters at P3, P4, and P5 boost accuracy by +0.954 pp over a matched plain B0 control (73.88% vs 72.92%).
  • Attention amplitude increases with depth: 27.2% at P3, 45.4% at P4, and 66.3% at P5.
  • Alternating two shared MBConv templates in Stage 6 reduces parameters by 10.9% while losing only 1.84 pp accuracy.
  • Post-hoc kernel averaging and single-template stage tying cause severe accuracy degradation.

Highlights and analysis are generated automatically and may contain errors. Check the original source.