Skip to content
AI News HubLIVE
Original source2 min read

Sketch, Don't Paint: Allocating Test-Time Compute in Generative Visual Reasoning

Summary

This paper studies how multimodal models should allocate test-time compute in visual reasoning: when to render visual thoughts and how many denoising steps each thought needs. It finds that visual thoughts fall into causal, decorative, and harmful regimes, that 8-step sketches often match 50-step renders, and proposes sketch-and-vote, which beats full rendering with fewer denoising steps across several reasoners.

SourcearXiv Computer VisionAuthor: Sheng Lu, Guangpin Tao, Tong Lu, Ying Tai
Sketch, Don't Paint: Allocating Test-Time Compute in Generative Visual Reasoning
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 16 Nov 2026]

Title:Sketch, Don't Paint: Allocating Test-Time Compute in Generative Visual Reasoning

View a PDF of the paper titled Sketch, Don't Paint: Allocating Test-Time Compute in Generative Visual Reasoning, by Sheng Lu and 3 other authors

View PDF HTML (experimental)

Abstract:Multimodal models increasingly reason by synthesizing pixels, sketching maze paths within interleaved chains of thought or rolling out video solutions frame by frame. Each visual thought runs the full diffusion schedule inherited from photorealistic synthesis, so iterative denoising dominates inference cost. We study how to allocate test-time compute along two axes: when to render visual thoughts rather than thinking in text or answering directly, and how many denoising steps each thought requires. Controlled interventions that vary only how the thought is rendered uncover three regimes. Causal thoughts carry the plan and become legible early: an 8-step sketch nearly matches a 50-step render in accuracy, commits to the same plan as often as two independent full renders agree, and in video models mostly fails by stalling short of the goal, so the rollout abstains. Decorative thoughts can be omitted without loss, whereas harmful thoughts lower accuracy. Because the regime depends on both model and task, always drawing is suboptimal. Across 14 model-benchmark pairs of three interleaved reasoners, answering directly, without any thinking, wins 9 pairs, and choosing among direct, text-only, and visual tiers from ten labeled examples per pair raises mean accuracy from 60.7% (always drawing) to 65.1%. Where drawing pays, we propose sketch-and-vote, which samples 8-step sketches (a step count chosen without labels) from independent noises until one answer reaches three votes. With this one setting, sketch-and-vote beats full rendering with fewer denoising steps on an image reasoner and two video reasoners, by 8.6 points on VSP spatial planning and by 9.6 and 22.1 points on mazes.

Comments: 11 pages, 5 figures, 3 tables

Subjects:

Computer Vision and Pattern Recognition (cs.CV)

Cite as: arXiv:2611.21812 [cs.CV]

(or arXiv:2611.21812v1 [cs.CV] for this version)

https://doi.org/10.48550/arXiv.2611.21812

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Guangpin Tao [view email] [v1] Mon, 16 Nov 2026 19:25:10 UTC (2,043 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Sketch, Don't Paint: Allocating Test-Time Compute in Generative Visual Reasoning, by Sheng Lu and 3 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

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

  • Causal visual thoughts become legible early, with 8-step sketches nearly matching 50-step renders in accuracy.
  • Direct answering wins 9 of 14 model-benchmark pairs, so always drawing is suboptimal.
  • Choosing among direct, text-only, and visual tiers from ten labeled examples per pair raises mean accuracy from 60.7% to 65.1%.
  • Sketch-and-vote samples 8-step sketches until one answer reaches three votes and beats full rendering with fewer denoising steps.

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

Sketch, Don't Paint: Allocating Test-Time Compute in Generative Visual Reasoning | AI News Hub