From World Models to World Action Models: A Concise Tutorial for Robotics
摘要
World models are increasingly used in embodied intelligence and generative simulation, yet their scope remains ambiguous across communities. This tutorial presents a design-space view of world models as action-conditioned predictive models that estimate the future evolution of task-relevant observations or states. We categorize existing methods into observation-space and state-space world models, comparing their trade-offs in visual fidelity, spatial structure, physical interpretability, and control usability. We further introduce world action models, which connect predicted futures with executable robot actions, and summarize four representative paradigms: imagine-then-execute, video-feature-conditioned action prediction, joint video-action modeling, and auxiliary video prediction for policy learning. The goal of this tutorial is to clarify the conceptual scope of world (action) models and provide a structured taxonomy for embodied prediction and control.
引用
@article{arxiv.2607.00836,
title = {From World Models to World Action Models: A Concise Tutorial for Robotics},
author = {Xiaoxiong Zhang and Xiong Zeng and Wei Zhang},
journal= {arXiv preprint arXiv:2607.00836},
year = {2026}
}
备注
Project page: https://clearlab-sustech.github.io/WorldModelSurvey/