WorldDraft: If Code Builds the World, Can Video Models Just Render It?
Summary
The paper introduces WorldDraft, a benchmark of 432 paired test cases across 48 executable 4D world families, evaluating whether video models can act as generative renderers of code-built worlds. It finds a stark tension between world preservation and visual realization across 12 models, and proposes WorldDraft-Primer, a plug-and-play visual prompting strategy that improves both objectives.
View Cached Full Text
Cached at: 11/16/26, 08:56 AM
Paper page - WorldDraft: If Code Builds the World, Can Video Models Just Render It?
Source: https://huggingface.co/papers/2611.16217
Abstract
Multimodalcodingagentsdecouple4Dworldconstructionfromvisualgeneration,repurposingvideomodelsfromworldgeneratorstowardgenerativerenderers.Thisdivisionoflaborraisesafundamentalquestion:ifcodebuildstheworld,canvideomodelsjustrenderit?Areliablegenerativerendererhingesontwocomplementarycriteria:(i)worldpreservation,retainingprescribed4Dkinematicsandcamerapaths;and(ii)visualrealization,persistentlybindingtext-specifiedtargetattributesovertime.Toevaluatethiscapability,weintroduceWorldDraft,abenchmarkof432strictlypairedtestcasesacross48executable4Dworldfamiliesspanningsixdynamicchallengecategories.Undera3times3factorialdesign,WorldDraftorthogonallycrossesdraftgranularity(abstractproxytoprevis)andappearancealignment(neutraltoconflicting)underidenticaldynamics.Extensiveevaluationsacross12representativemodelsrevealthat:(I)worldpreservationandvisualrealizationremaininstarkoperationaltension;(II)richerdraftscaffoldinganchorsstructurebutparadoxicallyimpairsappearancefidelity;and(III)conflictingsourcecuestriggerseveresourceleakage(upto64%),exposingfragilecross-modaldisentanglement.Tomitigatethesebottlenecks,weproposeWorldDraft-Primer,aplug-and-playvisualpromptingstrategythatconsistentlyimprovesbothobjectivesacrossfrozenvideorenderers.Together,WorldDraftandWorldDraft-Primerlayafoundationforbridgingprogrammatic4Dsimulationwithvideogeneration,advancingphysicallygroundedgenerativerendering.
View arXiv pageView PDFProject pageGitHub15Add to collection
Get this paper in your agent:
hf papers read 2611\.16217
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/2611.16217 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2611.16217 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2611.16217 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.
Similar Articles
Code World Model 2.0: Towards Infinite Controllable Video Generation
The paper presents Invoker-Render (Code World Model 2.0), pairing a bidirectional creation model with a causal four-step streaming model via Progressive Forcing to enable controllable video generation from structural proxies, references, and language, with stable 30-minute self-generated rollouts.
Wonder: Video World Model Done Better
Wonder is a general-purpose video world model that enables real-time, camera-controllable world exploration from an image or conditional video. It introduces camera conditioning via dense coordinate fields, a sparse attention memory mechanism, and techniques to improve distillation, allowing minute-scale video generation at 16 FPS.
VisualPatchWorld: Code World Models as Latent Structured Representations for Planning
VisualPatchWorld introduces a method for learning world dynamics as code, enabling inspectable and editable simulators from data. It achieves strong planning success in navigation and manipulation tasks.
Code World Model: Coding Agent as World Brain
The paper introduces Code World Model, a framework that combines language models for reasoning and coding with video models for visual rendering to simulate persistent world evolution in an open-ended manner.
MiniWorld: Democratizing the Training of Video World Models from Scratch
MiniWorld is a reproducible framework for training video world models from scratch using a block-causal Video Diffusion Transformer with Flow Matching, enabling efficient streaming generation and trainable in days on a single 8-GPU server.