From World Action Models to Embodied Brains: A Roadmap for Open-World Physical Intelligence

2026-07-13

AI Summary

Current research in physical intelligence is severely hindered by incompatible action spaces, inconsistent task benchmarks, and fragmented interfaces, impeding progress toward general-purpose embodied intelligence. This work proposes a co-evolutionary framework oriented toward the long-term vision of an “embodied brain,” offering the first systematic definition of this concept and introducing a modular physical intelligence stack. The framework leverages World Action Models (WAMs) for interventional prediction, aligns heterogeneous components through a physics-binding layer and shared interface contracts, and incorporates a closed-loop post-training mechanism that transforms validation interactions into reusable experience. Designed to support adaptability and self-improvement, this architecture provides a unified and scalable foundation for embodied agents operating in open-world environments.

Abstract

Artificial general intelligence ultimately requires agents that can reason and act in the physical world. Action models, vision-language-action policies, and world models have advanced this goal, while World Action Models (WAMs) are particularly promising because they connect candidate interventions with predicted consequences. However, progress remains fragmented: models use incompatible action spaces and prediction targets, datasets and tasks follow different conventions, and runtime systems expose limited interfaces for reuse and evaluation. We review the evolution toward WAMs and organize these limitations into three coupled gaps: model roles and representations, objectives and standardization, and system composition. Building on this analysis, we propose a co-evolution roadmap for physical intelligence centered on the \emph{embodied brain}, a long-term model target for integrating multimodal context, comparing candidate interventions, and issuing state-transition or capability requests rather than direct actuator commands. WAMs provide promising prototypes for its predictive functions, while a physical harness grounds model outputs through tools, controllers, verification, and trace logging. Shared contracts align heterogeneous models, data, tasks, and embodiments, and closed-loop post-training converts verified interaction into reusable experience. Together, these components define a modular physical-intelligence stack for adaptive and self-improving embodied agents.

Problem

• World Action Models
• physical intelligence
• embodied agents
• standardization
• action spaces

Innovation

• World Action Models
• embodied brain
• physical intelligence
• modular stack
• shared contracts