aiqbench analysis
This research introduces the Quadrupedal World Model (QWM), a generative dynamics model capable of zero-shot cross-embodiment transfer for quadruped robots by conditioning on scale-invariant physical features. Unlike traditional morphology-locked models that require retraining when robot dimensions change, QWM trains policies entirely within its latent space to generalize across different actuator and limb configurations without fine-tuning. This advancement represents a significant step toward scalable robotics where physics-based world models can efficiently adapt behavior for diverse hardware families once the environment dynamics are learned.
Morphology-Conditioned World Model for Cross-Embodiment Quadrupedal Locomotion
arXiv:2604.08780v2 Announce Type: replace-cross Abstract: World models promise a paradigm shift in robotics, where an agent learns the physics of its environment once and then acquires behaviors efficiently. Yet the learned dynamics models at their core are typically morphology locked. In legged locomotion, a dynamics model trained on an ANYmal-D quadruped fails on a Unitree Go1 because it overfits to one robot's embodiment rather than capturing the locomotion dynamics shared across robots, so even a small change in actuator dynamics or limb length forces retraining from scratch. However, if we formalize a robot's unique physical traits into a morphology specification, a controller for a family of robots can utilize this blueprint in two ways. It can feed the specification to a model-free policy, or it can feed the specification to a learned dynamics model and extract the policy in imagination. We argue for the second route and introduce the Quadrupedal World Model (QWM), which conditions a single generative dynamics model on scale-invariant physical features and trains policies entirely inside it, through a physical morphology encoder, an adaptive reward normalizer, and morphology conditioning in the latent dynamics. Holding the morphology information identical, a model-free policy matches QWM on the training cohort but degrades on unseen morphologies, while QWM transfers zero-shot with no fine-tuning, adaptation, or warm-up in such cases. To our knowledge, this is the first world model to demonstrate zero-shot cross-embodiment transfer within the quadrupedal family.
Abstract:World models promise a paradigm shift in robotics, where an agent learns the physics of its environment once and then acquires behaviors efficiently. Yet the learned dynamics models at their core are typically morphology locked. In legged locomotion, a dynamics model trained on an ANYmal-D quadruped fails on a Unitree Go1 because it overfits to one robot's embodiment rather than capturing the locomotion dynamics shared across robots, so even a small change in actuator dynamics or limb length forces retraining from scratch. However, if we formalize a robot's unique physical traits into a morphology specification, a controller for a family of robots can utilize this blueprint in two ways. It can feed the specification to a model-free policy, or it can feed the specification to a learned dynamics model and extract the policy in imagination. We argue for the second route and introduce the Quadrupedal World Model (QWM), which conditions a single generative dynamics model on scale-invariant physical features and trains policies entirely inside it, through a physical morphology encoder, an adaptive reward normalizer, and morphology conditioning in the latent dynamics. Holding the morphology information identical, a model-free policy matches QWM on the training cohort but degrades on unseen morphologies, while QWM transfers zero-shot with no fine-tuning, adaptation, or warm-up in such cases. To our knowledge, this is the first world model to demonstrate zero-shot cross-embodiment transfer within the quadrupedal family.
Robotics (cs.RO); Machine Learning (cs.LG)
arXiv:2604.08780 [cs.RO]
(or arXiv:2604.08780v2 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2604.08780 Focus to learn more arXiv-issued DOI via DataCite
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 · 325 words
Related coverage
CLAP: Cross-Embodiment Video World Models are Zero-Shot Physical Simulators
arXiv · Aug 28, 2026
Multi-Agent Egocentric World Model with Fine-Grained Embodied Interaction
arXiv · Oct 10, 2026
MorFiC: Fixing Value Miscalibration for Zero-Shot Quadruped Transfer
arXiv · Aug 25, 2026
Meet Qwen-RobotSuite: Three Embodied AI Models for VLA Manipulation, Video World Modeling, and Navigation
MarkTechPost · Jun 16, 2026