OpenWAM: An Open, Modular Exploration Towards Systematic World-Action Model Pretraining

Hugging Face Daily Papers Papers
world-action-models pretraining modular-design open-research robotics simulation benchmark

Summary

OpenWAM presents an open, modular framework for world-action model pretraining to systematically identify design principles, with pretrained models showing strong performance in simulation and real-world robotics.

World-Action Models inherit world knowledge from video-generative priors, and channel it into executable control signals through embodied experience. Existing systems, however, are monolithic: the generative backbone, visual representation, architecture, information flow, inference procedure, and training data are tightly coupled, obscuring which design choices matter and why. We introduce OpenWAM, an open research stack that turns world-action pretraining into a controlled experimental program. OpenWAM-Infra factorizes the WAM design space into composable modules with unified training, inference, deployment, and evaluation. On this substrate, OpenWAM-Study examines three questions through controlled experiments: what to inherit, how world and action learning interact, and how their synergy scales; and distills three principles: upstream knowledge transfers through a sufficiently capable generative backbone and a compact, information-rich latent space; world-action synergy requires dedicated action capacity, explicit world-to-action information flow, and synchronized joint denoising; and embodied pretraining principally improves out-of-domain generalization, with one-stage co-training over egocentric and robot data integrating world coverage and action grounding. Composing these principles, we build OpenWAM-α, an open WAM pretrained on roughly 6,400 hours of egocentric human and robot data and evaluated across simulation and real-world benchmarks. Across the eight simulation benchmarks and the real-robot experiments, which together span embodiments from single-arm and bimanual manipulation to dexterous hands, OpenWAM-α delivers consistently excellent performance, sustaining its top-tier standing from simulation to the physical world. We release the full stack, including infrastructure, evaluation protocols, pretrained models, and data recipes, to facilitate future research.
Original Article
View Cached Full Text

Cached at: 09/09/26, 12:31 PM

Paper page - OpenWAM: An Open, Modular Exploration Towards Systematic World-Action Model Pretraining

Source: https://huggingface.co/papers/2609.07398 Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

OpenWAM factorizes world-action pretraining into modular components to identify key design principles, yielding a scalable open model with strong simulation and real-robot performance.

World-Action Modelsinherit world knowledge fromvideo-generative priors, and channel it into executable control signals through embodied experience. Existing systems, however, are monolithic: thegenerative backbone, visual representation, architecture, information flow, inference procedure, and training data are tightly coupled, obscuring which design choices matter and why. We introduce OpenWAM, an open research stack that turns world-action pretraining into a controlled experimental program. OpenWAM-Infra factorizes the WAM design space into composable modules with unified training, inference, deployment, and evaluation. On this substrate, OpenWAM-Study examines three questions through controlled experiments: what to inherit, how world and action learning interact, and how their synergy scales; and distills three principles: upstream knowledge transfers through a sufficiently capablegenerative backboneand a compact, information-richlatent space; world-action synergy requires dedicated action capacity, explicit world-to-action information flow, and synchronizedjoint denoising; and embodied pretraining principally improves out-of-domain generalization, with one-stageco-trainingover egocentric and robot data integrating world coverage and action grounding. Composing these principles, we build OpenWAM-α, an open WAM pretrained on roughly 6,400 hours of egocentric human and robot data and evaluated across simulation and real-world benchmarks. Across the eight simulation benchmarks and the real-robot experiments, which together span embodiments from single-arm and bimanual manipulation to dexterous hands, OpenWAM-α delivers consistently excellent performance, sustaining its top-tier standing from simulation to the physical world. We release the full stack, including infrastructure, evaluation protocols, pretrained models, and data recipes, to facilitate future research.

View arXiv pageView PDFProject pageGitHub344Add to collection

Models citing this paper46

#### OpenWAM/OpenWAM-Alpha-Pretrain-Foundation-Model Robotics• Updatedabout 4 hours ago • 42 #### OpenWAM/OpenWAM-Alpha-Real-Dexterous-Hand-Wuji Robotics• Updatedabout 4 hours ago • 10 #### OpenWAM/OpenWAM-Alpha-Real-RoboDojo-ARX-X5 Robotics• Updatedabout 4 hours ago • 9 #### OpenWAM/OpenWAM-Alpha-Real-RoboDojo-Piper Robotics• Updatedabout 4 hours ago • 12 Browse 46 models citing this paper## Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2609.07398 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.07398 in a Space README.md to link it from this page.

Collections including this paper1

Similar Articles

World Action Modeling with Progressive Visual Planning

arXiv cs.AI

ProWAM is a progressive world action model that jointly predicts robot actions and an ordered sequence of sparse visual sub-goals, enabling efficient long-horizon control without dense video generation. It achieves state-of-the-art results on LIBERO-Plus and RoboTwin, plus 70% zero-shot real-world success.

World Action Models: The Next Frontier in Embodied AI

Hugging Face Daily Papers

This survey paper introduces World Action Models (WAMs), a unified framework for embodied AI that integrates predictive state modeling with action generation. It provides a taxonomy of existing methods, analyzes the data ecosystem, and outlines evaluation protocols for this emerging paradigm.

PointWAM: 3D World Action Modeling for Dexterous Robotic Manipulation

Hugging Face Daily Papers

PointWAM introduces a 3D world action model that decomposes the world into scene and hands, jointly forecasting both as 3D point trajectories to enable dexterous robotic manipulation. Pre-training on human demonstration videos boosts DexJoCo success by 56.9 points, surpassing prior state-of-the-art by 11.7 points and outperforming strong VLAs on a real robot.