Skip to content
AI News HubLIVE
Original source2 min read

Training-Free Acceleration of Multi-View 3D Transformers via Regime-Aware Attention Approximation

Summary

Researchers analyze global attention in multi-view 3D transformers like VGGT, finding each layer plays one of three roles, and propose RAWATTA, a training-free method using 8-bit attention with selective 16-bit precision to accelerate inference, fitting 200 images on a 12GB GPU with minimal accuracy loss.

SourcearXiv Computer VisionAuthor: Muhammed Izzet Sezer, Sevval Belkis Dikkaya
Training-Free Acceleration of Multi-View 3D Transformers via Regime-Aware Attention Approximation
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 13 Nov 2026]

Title:Training-Free Acceleration of Multi-View 3D Transformers via Regime-Aware Attention Approximation

View a PDF of the paper titled Training-Free Acceleration of Multi-View 3D Transformers via Regime-Aware Attention Approximation, by Muhammed Izzet Sezer and 1 other authors

View PDF HTML (experimental)

Abstract:Multi-view 3D transformers such as VGGT reconstruct a scene from a set of unposed images in a single forward pass, predicting camera poses, depth and 3D points, but the cost of their global attention grows quadratically with the num- ber of images. Accelerators borrowed from large language models assume that each token attends to only a few oth- ers; on these models they lose accuracy or gain little speed. We study the global attention of three multi-view transform- ers on seven datasets and find that every layer plays one of three roles: it averages over the whole scene, matches points across many images, or anchors on the first image, whose coordinates the outputs use. Building on this anal- ysis, we propose Regime-AWare ATTention Approximation (RAWATTA), a training-free method that runs attention in 8- bit precision and keeps 16-bit precision only in the layers where a worst-case calibration check requires it. Its fast set- ting, RAWATTA-Fast, also approximates attention in a few averaging and anchoring layers, while matching layers and the first image keep all their keys. On a 12 GB GPU at 50 im- ages, RAWATTA runs VGGT faster than the released code, mostly through our CUDA engine, and loses at most about one point of pose accuracy (AUC@3) on any dataset; RAWATTA-Fast is slightly faster and less accurate. Both fit 200 images, where the released code runs out of memory at 100.

Subjects:

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

Cite as: arXiv:2611.17629 [cs.CV]

(or arXiv:2611.17629v1 [cs.CV] for this version)

https://doi.org/10.48550/arXiv.2611.17629

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Muhammed Sezer [view email] [v1] Fri, 13 Nov 2026 00:59:07 UTC (1,382 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Training-Free Acceleration of Multi-View 3D Transformers via Regime-Aware Attention Approximation, by Muhammed Izzet Sezer and 1 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 cs.AI

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

EngineersAdvanced

Key points

  • Attention layers in multi-view 3D transformers fall into three functional roles: whole-scene averaging, cross-image point matching, and first-image anchoring.
  • RAWATTA runs attention in 8-bit precision and retains 16-bit only where a worst-case calibration check requires it, without retraining.
  • On a 12GB GPU at 50 images, RAWATTA is faster than released VGGT code with at most ~1 point AUC@3 pose accuracy loss.
  • RAWATTA-Fast is slightly faster but less accurate; both handle 200 images where the original code runs out of memory at 100.

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