Paper 2606.20781
World Action Models: A Survey
- Published
- Jun 2026
- Research lab
- Independent
- Citations
- 1
- GitHub
- 308 stars
01 In brief
Summary
This survey defines World Action Models (WAMs) as embodied predictive-action models that make a forecast of the future available to action, distinguishing them from VLAs, world models, and video generation models.
It organizes WAMs via two complementary views: a design-philosophy taxonomy (Render-and-Decode, Latent-Only, Video-Generation-Free) and a component-level anatomy with four axes (predictive substrate, architectural backbone, action coupling, deployment regime).
The survey discusses five core properties (interactability, causality, persistence, physical plausibility, generalization), data sources, evaluation practices, and open challenges.
A consistent design pattern emerges: WAMs are not simply video generators with action heads, but predictive-action methods trading representational richness against compute, memory, latency, and action-label cost.
The field is moving toward generating less of the future while preserving what control requires.
02 From the paper
Abstract
World Action Models (WAMs) are embodied predictive-action models that make a forecast of the future available to action. Recent WAMs repurpose large video generation models, and a parallel line relies on language or vision-language backbones without a video-generation core. This rapid expansion has blurred the boundary among broad world models, video generation models, action-grounded video world models, Vision-Language-Action policies, and WAMs. This survey gives the field a common account. It first clarifies these boundaries, then organizes existing works through two complementary views. The first view asks what each method is required to generate, spanning rendered futures, latent futures, and video-generation-free action reasoning. The second view decomposes each method by predictive substrate, backbone, action coupling, and deployment regime. This anatomy supports a unified discussion of interactability, causality, persistence, physical plausibility, and generalization, followed by data, evaluation, and open challenges. Across these axes, a consistent design pattern emerges: WAMs are not simply video generators with action heads, but predictive-action methods whose design choices trade representational richness against compute, memory, latency, and action-label cost. The field is moving toward methods that generate less of the future while preserving what control requires. The survey homepage is available at https://world-action-models.github.io/.