aiqbench analysis
GeniWorld advances robotic manipulation by leveraging pretrained video generative models to create an interactive world model capable of robust zero-shot generalization in unseen environments. By decoupling robot kinematics from environmental dynamics, the system enables closed-loop control and scalable policy evaluation without requiring extensive real-world data collection for every new scenario. This approach directly addresses critical limitations in current generalist policies regarding out-of-distribution scenarios and complex environment adaptation.
GeniWorld: A Generalizable Interactive World Model for Robotic Manipulation via Visual Actions
arXiv:2608.06332v1 Announce Type: new Abstract: Generalist robot policies exhibit strong capabilities, but their robustness in complex and unseen environments remains limited. Scaling robot learning and evaluation in diverse real-world environments remains costly and challenging. Action-conditioned world models offer a promising alternative, but they often suffer from limited action controllability and poor generalization to out-of-distribution (OOD) scenarios. To this end, we present GeniWorld, an interactive world model for robots that generalizes robustly across unseen scenarios. Building on pretrained video generative models, we use URDF-based rendering to transform numerical actions into visual action representations, enabling spatially grounded action control. By explicitly decoupling embodiment kinematics from environmental dynamics, our model mitigates scene overfitting and facilitates modeling of robot-environment interactions. To achieve closed-loop control, we construct an autoregressive video prediction model integrated with high-frequency robot kinematic control, enabling interaction with both robot policies and human teleoperators. In our experiments, even when trained solely on limited fixed-scene data, our model achieves superior in-domain performance and robust zero-shot generalization to highly randomized, unseen environments. For downstream applications, GeniWorld serves as a scalable policy evaluator that remains reliable under environmental perturbations. Furthermore, even with limited real-world demonstrations, GeniWorld generates diverse manipulation trajectories within the world model, improving downstream policy performance and robustness in complex environments.
Abstract:Generalist robot policies exhibit strong capabilities, but their robustness in complex and unseen environments remains limited. Scaling robot learning and evaluation in diverse real-world environments remains costly and challenging. Action-conditioned world models offer a promising alternative, but they often suffer from limited action controllability and poor generalization to out-of-distribution (OOD) scenarios. To this end, we present GeniWorld, an interactive world model for robots that generalizes robustly across unseen scenarios. Building on pretrained video generative models, we use URDF-based rendering to transform numerical actions into visual action representations, enabling spatially grounded action control. By explicitly decoupling embodiment kinematics from environmental dynamics, our model mitigates scene overfitting and facilitates modeling of robot-environment interactions. To achieve closed-loop control, we construct an autoregressive video prediction model integrated with high-frequency robot kinematic control, enabling interaction with both robot policies and human teleoperators. In our experiments, even when trained solely on limited fixed-scene data, our model achieves superior in-domain performance and robust zero-shot generalization to highly randomized, unseen environments. For downstream applications, GeniWorld serves as a scalable policy evaluator that remains reliable under environmental perturbations. Furthermore, even with limited real-world demonstrations, GeniWorld generates diverse manipulation trajectories within the world model, improving downstream policy performance and robustness in complex environments.
Robotics (cs.RO)
arXiv:2608.06332 [cs.RO]
(or arXiv:2608.06332v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2608.06332 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
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.
Full text under the publisher's open licence · 296 words
Related coverage
Riemann-1.0: An Embodied World Action Model for Physical AI
arXiv · Aug 28, 2026
AffordanceWAM: Affordance-Aware Joint World-Action Modeling for Robot Manipulation
arXiv · Sep 22, 2026
AnyWorld: Factorized Egocentric World Models for Cross-Embodiment Generalization
arXiv · Sep 1, 2026
HERO: Learning Humanoid End-Effector Control for Visual Whole-Body Open-Vocabulary Object Grasping
arXiv · Oct 15, 2026