Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

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

Source

arXiv

arxiv.org/abs/2607.11689v1

Read the source

Source

Plain-Language Brief

The paper proposes a roadmap for developing artificial general intelligence that can reason and act in the physical world. The authors identify three gaps in current progress: inconsistencies in model roles, objectives, and standards; lack of system composition. They suggest an embodied brain model as a long-term goal to integrate multimodal context, compare interventions, and issue state-transition requests. The roadmap includes using World Action Models for prediction functions, physical harness for grounding outputs, shared contracts for aligning models and data, and closed-loop post-training for reusable experience.

What the Document Says

  • Artificial general intelligence requires agents to reason and act in the physical world.
  • Action models, vision-language-action policies, and world models are advancing this goal.
  • World Action Models (WAMs) connect interventions with consequences.
  • Progress is fragmented due to incompatible action spaces, prediction targets, inconsistent datasets and tasks, and limited runtime system interfaces.
  • The authors identify three gaps: model roles and representations, objectives and standardization, and system composition.
  • They propose the embodied brain as a long-term model target for integrating multimodal context, comparing interventions, and issuing state-transition requests.
  • WAMs provide promising prototypes for predictive functions.
  • A physical harness grounds model outputs through tools, controllers, verification, and trace logging.
  • Shared contracts align heterogeneous models, data, tasks, and embodiments.
  • Closed-loop post-training converts verified interaction into reusable experience.
  • This defines a modular physical-intelligence stack for adaptive and self-improving embodied agents.

Key Concepts Mentioned

  • Artificial General Intelligence
  • World Action Models (WAMs)
  • Embodied Brain
  • Physical Harness
  • Shared Contracts
  • Closed-Loop Post-Training

Claims Made by the Document

  • Current progress in artificial general intelligence is fragmented.
  • The embodied brain model can integrate multimodal context, compare interventions, and issue state-transition requests.
  • WAMs are promising prototypes for the predictive functions of the embodied brain.

Evidence Used

  • The text discusses inconsistencies in current progress and proposes a solution with the embodied brain.

Methods / Approach

  • Review of the evolution towards World Action Models (WAMs) and identification of limitations.
  • Proposal of a co-evolution roadmap for physical intelligence centered on the embodied brain.

Limitations Stated by the Authors

  • The paper does not provide specific examples or implementations of the proposed embodied brain model.

Limitations Noted by the Agent

Agent interpretation — not from the source.

  • The text does not specify how the proposed roadmap will be implemented or evaluated.
  • It is unclear how the shared contracts and closed-loop post-training components will be achieved in practice.

Open Questions

  • How can the embodied brain model be implemented and evaluated?
  • What specific methods are used for shared contracts and closed-loop post-training?

Notes for Future Q&A

May include interpretation. Labeled where it does.

  • The paper discusses a roadmap for physical intelligence, but does not provide concrete details about how it will be implemented or tested.

Raw Abstract / Excerpt

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.