Skip to content
AI News HubLIVE
Original source2 min read

Bits, Not Boxes: Label-Free Layer Selection for Visual Localization in VLMs

Summary

WhereBits identifies which layers in vision-language models localize queried objects by measuring the image-independent component of attention in bits on unlabeled questions, eliminating the need for box annotations. Across eight VLMs, it matches box-tuned pointing accuracy and improves V* accuracy by 6.0 points in training-free zoom-in.

SourcearXiv Computer VisionAuthor: Haisheng Fu, Gexin Huang, Haikai Zhao, Jie Liang, Zhenman Fang, Jingning Han, Xiaoxiao Li
Bits, Not Boxes: Label-Free Layer Selection for Visual Localization in VLMs
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:Bits, Not Boxes: Label-Free Layer Selection for Visual Localization in VLMs

View a PDF of the paper titled Bits, Not Boxes: Label-Free Layer Selection for Visual Localization in VLMs, by Haisheng Fu and 6 other authors

View PDF HTML (experimental)

Abstract:Attention maps cheaply show where a vision-language model (VLM) looks, but attention sinks make them unreliable for localization. Contrasting the question with a generic prompt suppresses sinks, yet only some layers localize the queried object. Existing methods find these layers with box annotations or fixed rules. We show that attention itself can identify them. A question changes how much a head attends to the image and where this attention goes. The two effects separate exactly, and the second, measured in bits on unlabeled questions, indicates which layers localize. WhereBits turns these where-to-look bits into a layer prior and combines it with per-head contrastive scores. Its three design constants were fixed once on labeled COCO data, and a new model or resolution needs only unlabeled questions. Across eight VLMs, WhereBits matches the pointing accuracy of box-tuned layers on COCO and RefCOCO and ranks target tokens better. This advantage holds on a held-out model family and at higher resolutions. Its COCO pointing accuracy is 0.72, against 0.22 for mean attention. In training-free zoom-in, its crops raise V* accuracy by 6.0 points under forced-choice evaluation.

Comments: 35 pages, 9 figures, 56 tables

Subjects:

Computer Vision and Pattern Recognition (cs.CV); Methodology (stat.ME)

Cite as: arXiv:2611.23432 [cs.CV]

(or arXiv:2611.23432v1 [cs.CV] for this version)

https://doi.org/10.48550/arXiv.2611.23432

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Haisheng Fu [view email] [v1] Tue, 17 Nov 2026 19:37:06 UTC (7,701 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Bits, Not Boxes: Label-Free Layer Selection for Visual Localization in VLMs, by Haisheng Fu and 6 other authors

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 stat stat.ME

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

InvestorsAdvanced

Key points

  • The 'where-to-look' information in attention can be separated exactly from overall attention magnitude.
  • A layer prior derived from this bit measure, combined with per-head contrastive scores, localizes targets without annotations.
  • Matches box-tuned layer pointing accuracy on COCO and RefCOCO across eight VLMs and generalizes to new model families and resolutions.
  • In training-free zoom-in, its crops raise V* accuracy by 6.0 points under forced-choice evaluation.

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