<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> This repository serves as a curated, continuously updated reading list and resource hub for World Action Models (WAMs) in embodied AI. It addresses the need for a structured overview of this rapidly advancing field, targeting researchers and practitioners by consolidating key papers, blogs, and resources. The primary benefit is accelerating understanding and adoption of WAMs through a systematic survey and community-driven updates.

How It Works

The project centers on the paradigm of World Action Models (WAMs), which unify predictive world modeling with action generation for embodied AI agents. It categorizes WAMs into "Cascaded" and "Joint" architectures, detailing their core approaches, such as pixel-space representations and latent planning for Cascaded WAMs, and autoregressive or diffusion-based generation for Joint WAMs. This systematic classification and curated resource collection aim to provide a clear landscape of the field's advancements.

Quick Start & Requirements

This repository is a curated list of resources, not a runnable software project.

• Primary Install/Run: N/A. Users are directed to individual papers and projects for implementation.
• Prerequisites: Specific requirements depend on the individual WAM projects linked within the repository.
• Links:
  • Survey Paper (arXiv): https://arxiv.org/abs/2605.12090
  • Hugging Face: https://huggingface.co/papers/2605.12090
  • Project Website: https://openmoss.github.io/Awesome-WAM/

Highlighted Details

• Presents the first systematic survey on World Action Models (WAMs) for embodied AI.
• Features a comprehensive taxonomy of WAM architectures (Cascaded, Joint), training data, evaluation protocols, and world models for Vision-Language-Action (VLA) learning.
• Includes concise, structured summary blogs for each paper, with the summarization skill itself open-sourced.
• Employs a community-driven approach, actively soliciting contributions for missing papers and suggestions.

Maintenance & Community

The repository is designed to be "continuously updated as the field evolves," with an initial release on May 13, 2026. It explicitly encourages community involvement through issues and pull requests for additions and suggestions.

Licensing & Compatibility

No specific license information is provided in the README content.

Limitations & Caveats

As a curated list, this repository does not offer a unified framework or executable code for WAMs; users must consult individual linked resources for implementation details. The rapid evolution of the WAM field means the content requires ongoing updates to remain comprehensive.