Skip to content
AI News HubLIVE
Original source2 min read

Why Sub-4-Bit Vision CNNs Collapse and Where Mixed-Precision Recovery Stops

Summary

An arXiv study of ResNet and EfficientNet backbones on the full ImageNet-1k validation set maps sub-4-bit quantization collapse, showing degradation is highly non-linear and concentrated in late-stage pointwise projections and depthwise layers. Keeping under 5% of critical activations in 8-bit recovers up to 92% of lost accuracy, but software bit-unpacking causes 1.48x to 21.5x slowdowns on ARM and x86 CPUs.

SourcearXiv AIAuthor: Michael Schroeder
Why Sub-4-Bit Vision CNNs Collapse and Where Mixed-Precision Recovery Stops
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 (v1), last revised 19 Nov 2026 (this version, v2)]

Title:Why Sub-4-Bit Vision CNNs Collapse and Where Mixed-Precision Recovery Stops

View a PDF of the paper titled Why Sub-4-Bit Vision CNNs Collapse and Where Mixed-Precision Recovery Stops, by Michael Schroeder

View PDF HTML (experimental)

Abstract:Deep vision models operating in resource-constrained environments depend heavily on low-bit quantization, yet compressing weights below 4 bits frequently induces catastrophic accuracy collapse. While mixed-precision recovery strategies aim to rescue accuracy by selectively retaining high-precision parameters, the fundamental mechanics of where and why these models collapse remain poorly understood. In this study, we systematically analyze the anatomy of quantization collapse across standard Vision CNNs (including ResNet and EfficientNet backbones) evaluated on the full 50,000-image ImageNet-1k validation set. First, we map the exact progression of degradation, demonstrating that collapse is highly non-linear and concentrated in late-stage pointwise projections and depthwise layers. Second, we introduce a structured, multi-step recovery waterfall that evaluates selective integer precision (INT4/INT8) and selective floating-point scales. We show that keeping less than 5% of critical layer activations in 8-bit precision recovers up to 92% of the accuracy lost during aggressive sub-4-bit compression. Finally, we bridge the hardware reality gap by evaluating deployment performance on standard ARM and x86 CPU architectures. We show that while low-bit quantization reduces memory storage requirements, standard runtime software bit-unpacking introduces a 1.48x to 21.5x execution slowdown depending on batch size, highlighting a critical trade-off between model footprint and actual edge throughput. Our findings provide a practical blueprint for navigating the limits of sub-4-bit deployment without sacrificing hardware performance.

Comments: 8 pages, 6 figures, 2 tables

Subjects:

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

Cite as: arXiv:2611.23445 [cs.AI]

(or arXiv:2611.23445v2 [cs.AI] for this version)

https://doi.org/10.48550/arXiv.2611.23445

arXiv-issued DOI via DataCite

Submission history

From: Michael Schroeder [view email] [v1] Tue, 17 Nov 2026 19:49:27 UTC (77 KB)

[v2] Thu, 19 Nov 2026 14:30:24 UTC (77 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Why Sub-4-Bit Vision CNNs Collapse and Where Mixed-Precision Recovery Stops, by Michael Schroeder

View PDF

HTML (experimental)

TeX Source

view license

Additional Features

Audio Summary

Current browse context:

cs.AI

new | recent | 2026-11

Change to browse by:

cs cs.CV 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

  • Quantization collapse is highly non-linear and concentrated in late-stage pointwise projections and depthwise layers.
  • Holding less than 5% of critical layer activations in 8-bit recovers up to 92% of the accuracy lost.
  • Standard runtime software bit-unpacking slows execution by 1.48x to 21.5x on ARM and x86 CPUs.
  • There is a critical trade-off between reduced model footprint and actual edge throughput.

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