World Models: A Comprehensive Survey of Architectures, Methodologies, Reasoning Paradigms, and Applications

2026-05-28

AI Summary

This work addresses the lack of a unified framework in world model research, which has hindered systematic integration across diverse architectures, methodologies, reasoning paradigms, and applications. To bridge this gap, the paper proposes a four-dimensional taxonomy encompassing architectural designs (e.g., state-space models, Transformers, diffusion models, physics-informed networks), methodological families (e.g., language-augmented multimodal systems), reasoning paradigms (e.g., imagination-based planning and counterfactual reasoning), and application domains. For the first time, this framework connects foundational insights from cognitive science to the latest advances in large-scale models, revealing an emerging trend toward integrating chain-of-thought reasoning with world model imagination. It clarifies the overall developmental landscape, identifies core challenges such as error accumulation in prediction and simulation-to-reality transfer, and outlines a pathway toward unified multimodal world models.

Abstract

World models, internal simulators that learn the structure and dynamics of an environment, have emerged as a central paradigm in the pursuit of artificial general intelligence, enabling agents to predict, plan, and reason within learned representations. Despite rapid progress across reinforcement learning, robotics, autonomous driving, and video generation, the field lacks a unified framework integrating its diverse architectural choices, training methods, reasoning mechanisms, and application settings. This survey addresses that gap with a multi-axis taxonomy organized along four dimensions: (i) architecture, encompassing representation format, dynamics formulation, input modality, learning paradigm, and downstream application; (ii) methodological family, including state-space and recurrent approaches, transformer-based models, diffusion-based generators, physics-informed networks, and language-augmented multimodal systems; (iii) reasoning strategy, covering imagination-based planning, latent policy learning, counterfactual reasoning, and planning under uncertainty; and (iv) application domain, spanning robotics, autonomous driving, video prediction, multimodal agents, reinforcement learning, scientific modeling, medical imaging, educational measurement, and business and finance. Tracing the field from early cognitive-science foundations to milestone systems such as PlaNet, the Dreamer family, MuZero, Sora, Cosmos, and Genie, we examine how these dimensions interact and highlight the recent convergence of chain-of-thought reasoning with world-model imagination. We review evaluation protocols and benchmarks, identify persistent challenges such as compounding prediction errors, sim-to-real transfer, and fragmented evaluation, and outline future directions toward unified multimodal world models, foundation-scale interactive simulators, and safe deployment in safety-critical domains.

Problem

• world models
• unified framework
• architectural diversity
• reasoning paradigms
• application fragmentation

Innovation

• world models
• multi-axis taxonomy
• reasoning paradigms
• multimodal integration
• foundation-scale simulators