WorldDiT: A Unified Diffusion Architecture for World and Action Modeling
Summary
WorldDiT is a unified diffusion transformer architecture that couples action generation with visual world modeling, achieving strong performance on LIBERO simulation suites without relying on large pretrained vision-language models.
View Cached Full Text
Cached at: 07/28/26, 06:26 PM
Paper page - WorldDiT: A Unified Diffusion Architecture for World and Action Modeling
Source: https://huggingface.co/papers/2607.23909
Abstract
Manyrecentrobotpoliciespursuestrongercontrolbyusinglargepretrainedvision-languagemodels(VLMs)astheactionbackbone.WeintroduceWorldDiT,aunifieddiffusiontransformerarchitecturethatcouplesactiongenerationwithvisualworldmodelingandachievesstrongperformancewithoutalargepretrainedVLMactionbackbone.Duringtraining,asinglediffusiontransformergeneratescontinuousactionchunksandpredictsnormalizedRGBpatchtargetsfromfuturecameraframes.AcrossfourLIBEROsimulationsuites,WorldDiTliesonthereportedParetofrontierfortotalmodelparametersandmeansuccessamongmethodsreportingallfoursuites.Theseresultsprovideastrongsub-billion-parameterbaselineforfuturescalingstudies.
View arXiv pageView PDFAdd to collection
Community
Paper submitter
Upload images, audio, and videos by dragging in the text input, pasting, orclicking here.
Tap or paste here to upload images
Get this paper in your agent:
hf papers read 2607\.23909
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper1
#### bageldotcom/worlddit Robotics• Updatedabout 17 hours ago • 16 • 2
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2607.23909 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2607.23909 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
@bageldotcom: We are releasing WorldDiT, a unified architecture for robotics world modeling and control. On the LIBERO benchmark, it …
WorldDiT is a unified architecture for robotics world modeling and control, achieving the best performance on the LIBERO benchmark among methods that do not rely on a VLM for action generation, and lies on the reported Pareto frontier.
DWM: Separating World Effects from Actions in Latent World Models
Introduces DWM, a framework that decomposes latent world model transitions into action-driven and action-invariant (world effect) components, improving planning success on benchmarks with persistent world effects.
UniDDT: Unifying Multimodal Understanding and Generation with Decoupled Diffusion Transformer
UniDDT proposes a decoupled diffusion transformer framework that unifies multimodal understanding and generation by leveraging a Noisy ViT encoder and LLM for semantic encoding, achieving strong performance on both tasks.
Discrete-WAM: Unified Discrete Vision-Action Token Editing for World-Policy Learning
Introduces Discrete-WAM, a unified discrete latent vision-action world policy that enables compositional causal reasoning and counterfactual reasoning in autonomous driving through aligned discrete tokens and a shared discrete diffusion framework.
Flash-WAM: Modality-Aware Distillation for World Action Models
Flash-WAM introduces a modality-aware distillation method for world-action models, achieving real-time inference by compressing diffusion to a single step per modality, resulting in 23x speedup.