Paper page - How Far Can Coding Agents Go in Visual Generation?

Source: https://huggingface.co/papers/2610.19025

Abstract

Coding agents build images via executable programs, showing structured visual reasoning and consistent internal scene representations across drawing interfaces.

Modern image generation largely relies on specializedgenerative modelsthat leave the process of image construction implicit in learned parameters. We explore an alternative: general-purposecoding agentsthat construct images throughexecutable programs. Our study of zero-shot drawing onAgentDrawBench, a 200-prompt benchmark, reveals a distinctive capability profile. Agents excel at preserving structure and satisfying precisevisual constraints, achieving competitive results in domains where compositional accuracy matters more than photorealistic appearance, while remaining limited in perceptual richness. Their visual expressiveness varies with thedrawing interface, and a qualitative study on reconstruction shows howprocedural skillsextracted from human demonstrations support different artistic media. We then examine code as an externalvisual language of thought: a structured, executable representation that agents can construct and inspect. We test it invisual question answeringtask and find reconstructed visual code retains substantial information for question answering and, in some domains, supports higher accuracy than the original image. These abilities raise a further question: do agents draw on a shared internal representation of scene content beyond a particular drawing program? We explore aninternal visual language-of-thoughthypothesis, in which a shared, compositionally structured representation supports visual operations across drawing languages. Cross-harness drawings and images assembled from independent single-pixel queries exhibit related layouts and tones. Agents also maintain self-generated room arrangements across viewpoints after visible drawings are removed. These results demonstrate that the coding agent would maintain a consistent and coherent world inside its imagine even before the harness gets involved and the generation starts.

View arXiv pageView PDFProject pageAdd to collection

Get this paper in your agent:

hf papers read 2610\.19025

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2610.19025 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2610.19025 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2610.19025 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.