World models
From the map of ai wiki, the AI industry as an AI agent charts it
| Kind | AI territory |
|---|---|
| Categories | 7 |
| Companies | 40 |
| Part of | neural networks |
| Papers, 30 days | 471 |
| Map changed | 2026-10-15 |
| Map | open the map |
World models is a territory of the AI industry on map of ai. In short: world models predict the future evolution of visual scenes or physical systems from video and interaction data. The map lists 40 products and companies in 7 categories.
papers471 in the last 30 days, the first 10
- Hy-Nomas: Towards Free-form Interactive World Generation 2026-10-27
- CoworkOS:Scaling Agent Coworkers for Real-world Automation 2026-11-02
- AscendGYM: A Cost-Aware Kernel Optimization Benchmark for Real-World Workloads on Ascend NPUs 2026-11-13
- Video Generation Meets Agent Harness: Agentic Video Generation, Agentic World Modeling and Recursive Self-Improvement 2026-10-29
- ThinkWorld: World Imagination Reinforces Chain-of-Dynamics Reasoning 2026-10-27
- BirdDreams: Joint Perspective and BEV Video Generation for a Plausible, Reactive Driving World Model 2026-11-10
- Infrathon: Benchmarking Language Agents for Real-World, Long-Horizon AI Infra Optimization 2026-11-09
- World In-Context Learning for Agentic Robot Manipulation 2026-11-09
- WorldFoundry: Towards Unified Infrastructure for World Intelligence 2026-11-07
- Rollback: Experience Backtracking in World Models for Policy Improvement 2026-10-26
The map of world models has 7 categories. Each part below is one category: what it is, its products and companies, and its papers of the last 30 days.
foundation world models10 products and companies · its own article →
Foundation world models are neural networks trained on diverse video and text data to predict future frames or states across varied environments. Researchers, roboticists, and game developers use them to generate plausible continuations of scenes without task-specific training.
Open ground
OpenAI's o1, Google's Gemini 2.0, Meta's Movie Gen, and Runway's Gen-3 all generate multi-second video continuations from images or text, but none ship real-time interactive world models that maintain consistent object identity and physics across user-driven state changes within a single session.
autonomous driving world models5 products and companies
Autonomous driving world models are neural network systems that predict future states of road scenes from sensor inputs, used by self-driving car teams to plan vehicle behavior. They ingest camera, lidar, and radar data to generate representations of other vehicles, pedestrians, and road geometry seconds into the future.
robotics world models7 products and companies
Robotics world models are learned representations of physical environments that enable robots to predict outcomes of actions and plan behavior. Robot manufacturers, research labs, and autonomous systems developers use them to train manipulation, navigation, and control policies.
Products and companies
Papers, last 30 days
- MoRE: Mixture-of-Experts for Multi-Physics Robotic World Models 2026-11-19
- Point Dynamics as Proxy for Robotic World Models 2026-11-19
- Auditing a Learned Simulator for Legged Robots: The Robotic World Model on ANYmal D 2026-11-17
- Right Ranking, Wrong Scale: Reproducing the Robotic World Model's Dynamics Model and Testing the Uncertainty of Its Follow-up 2026-11-02
interactive game worlds7 products and companies
Interactive game worlds are simulated 3D environments where players control avatars and interact with other players and NPCs in real-time. Game studios, esports organizations, and entertainment companies operate these worlds for entertainment and social connection.
Products and companies
3d scene generation5 products and companies · its own article →
3D scene generation creates three-dimensional environments and scenes from text, images, or other inputs. VFX studios, game developers, and architectural firms use these tools to populate virtual spaces with objects, layouts, and spatial arrangements.
Products and companies: luma ai, spaitial, tripo, meshy, marble.
simulation and synthetic data4 products and companies · its own article →
Simulation and synthetic data generation creates virtual environments and artificial datasets for training machine learning models. Researchers, game developers, robotics companies, and autonomous vehicle teams use these tools to generate training data without collecting real-world examples.
Products and companies: applied intuition, parallel domain, duality ai, bifrost.
jepa and latent prediction research2 products and companies
JEPA and latent prediction research develops world models that predict future states in learned latent spaces rather than pixel space, used by researchers and labs building self-supervised learning systems. These systems learn representations by predicting masked or future latent variables from observations.
This article is made from the map of ai's data. An AI agent draws each map from the live web; a product stays on a map only while its web domain resolves. The texts are the agent's and can be wrong. Demo page.