A Definition and Roadmap for World Models

2026-07-07

AI Summary

This work addresses the lack of a unified definition, predictive objectives, and construction methodologies for “world models” in contemporary AI research—a gap that hinders cross-subfield collaboration. The paper proposes, for the first time, a consensus-based scientific definition of world models, systematically delineating their core components, capability boundaries, and essential technical elements. By integrating principles from dynamical systems modeling, representation learning, and multi-domain AI system design, it formulates a phased development roadmap. This framework offers a cohesive theoretical foundation and practical guidance for diverse areas such as reinforcement learning, video generation, and embodied intelligence, effectively bridging conceptual divides and advancing the systematic evolution of world models.

Abstract

World models -- internal simulators that learn the structure and dynamics of an environment -- have become one of the most actively debated concepts in AI. From model-based reinforcement learning and video generation to embodied robotics and ultimately, physical AI, researchers across AI subfields are building systems that they call "world models", yet there is no consensus on what a world model fundamentally is, what it should predict, or how it should be built. This perspective article provides a scientific definition of world models, discussions of their key technical aspects, and a staged roadmap for developing effective world models.

Problem

• world models
• definition
• consensus
• AI
• conceptual clarity

Innovation

• world models
• scientific definition
• roadmap
• model-based reinforcement learning
• embodied AI