DravenALG/awesome-vla-wam

What it solves

This project provides a centralized, organized directory of research into robotics foundation models. It helps researchers and developers navigate the rapidly evolving landscape of how robots perceive the world, understand language, and execute physical actions by categorizing complex academic approaches into clear, defined fields.

How it works

The repository acts as a curated knowledge base. It defines and groups various research methodologies—such as Vision-Language-Action (VLA) models, World Action Models (WAM), and Robotic Agentics—and links them to their foundational papers, websites, and technical reports. By establishing these definitions, it clarifies the relationships between different approaches to building intelligent, embodied robotic systems.

Who it’s for

This resource is designed for robotics researchers, AI engineers, and students who want to stay current with the state-of-the-art in embodied AI and foundation models for robotics. It is particularly useful for those looking to understand the technical distinctions between different policy-learning paradigms.

Highlights

• A systematic taxonomy of robotics research, including VLA, WAM, and Latent Action Models (LAM).
• A comprehensive, frequently updated list of academic papers and project websites for each category.
• Clear definitions for specialized terms like "Robotic Agentics" and "Human Action Pretraining" (HAP).
• Links to essential supporting resources, including robotics datasets, benchmarks, physics engines, and hardware documentation.