Top World Model Researchers in AI (2026)

Professor of EECS, UC Berkeley · Amazon Scholar / Head of LLM efforts in AGI org, Amazon

PhD · Stanford University — Andrew Ng

Berkeley leader in deep RL and robot learning. Co-founded Covariant. Now heads Amazon’s LLM and AGI robotics efforts.

Pioneered deep RL and imitation learning. He drove UniSim and Large World Model. Recent generalist robot world models include DreamDojo, trained on tens of thousands of hours of human video.

Associate Professor of Computer Science & EE, Stanford University · Co-Founder, Physical Intelligence

PhD · UC Berkeley — Sergey Levine and Pieter Abbeel

Co-created MAML meta-learning. Co-founded Physical Intelligence, which develops robot foundation models.

Pioneered meta-learning and visual foresight for robots. She coauthored Physical Intelligence’s pi-0.6, which uses RL self-improvement for VLAs. She also coauthored Ctrl-World and VLAW, which use video world models to evaluate and improve robot policies.

Co-founder & CEO, Sakana AI

PhD · University of Tokyo

Coauthored the foundational World Models paper. Co-founder and CEO of Sakana AI, which develops nature-inspired foundation models.

Coauthored the foundational 2018 World Models paper. It introduced the modern approach of learning compact predictive models of an environment for control. He now works at Sakana AI.

Co-founder, Embo

PhD · University of Toronto — Advised by Jimmy Ba (Vector Institute / U. Toronto)

Created the Dreamer world-model agents, including DreamerV3 and Dreamer 4. Develops RL through imagined interactions.

Created Dreamer for model-based RL. Dreamer 4 (2025) was the first agent to obtain diamonds in Minecraft using only offline data, without environment interaction. It marked a major advance in training agents within imagined world models.

Associate Professor of EECS (tenured), MIT · Distinguished Scientist (part-time), Google DeepMind

PhD · The Chinese University of Hong Kong — Xiaoou Tang

Lead author of ResNet for deep residual learning. Also contributed Faster R-CNN, Mask R-CNN, and MAE.

Created ResNet and Masked Autoencoders (MAE), among the most-cited works in deep learning. Recent work on JiT and improved Mean Flows advances diffusion-style models toward single-step generation without tokenizers. These methods provide core infrastructure for video world models.

Founder & Executive Chairman, AMI Labs (Advanced Machine Intelligence) · Professor, New York University

PhD · Université Pierre et Marie Curie (Sorbonne / UPMC), Paris — PhD on backpropagation; later postdoc with Geoffrey Hinton at U. Toronto

Turing Award winner who pioneered CNNs and deep learning. Promotes JEPA world-model architectures.

Turing Award winner who originated the Joint-Embedding Predictive Architecture (JEPA), the leading self-supervised alternative to autoregressive world models. His team’s 2025–26 work extended JEPA into a deployable family. It includes VL-JEPA, Causal-JEPA, and latent-action world models learned from video captured in everyday settings.

Co-founder & CEO, World Labs · Sequoia Professor of Computer Science; Co-Director, HAI, Stanford University

PhD · California Institute of Technology (Caltech) — Advised by Pietro Perona (and Christof Koch)

Created ImageNet. Co-founder and CEO of World Labs, which builds spatial-intelligence world models, including Marble.

Created ImageNet and leads the development of spatial intelligence. She co-founded World Labs, which raised about $1 billion. The company released Marble, which generates persistent 3D worlds from text and images, and the World API. It also released interactive world models that run in real time on a single GPU.

Co-founder & CEO, Recursive (Recursive Superintelligence) · Professor of Artificial Intelligence, University College London

PhD · University College London (UCL) — Advised by Sebastian Riedel

Led Genie foundation world models and Open-Endedness at DeepMind. UCL AI professor and Recursive co-founder.

Leads interactive world-model research at Google DeepMind. This includes Genie foundation world models, which generate playable, controllable environments from a single prompt.

Research Scientist, Meta FAIR

PhD · McGill University — Mila (Quebec AI Institute)

Lead author of I-JEPA, V-JEPA, and V-JEPA 2 self-supervised world-model architectures at Meta FAIR.

Lead author of I-JEPA and V-JEPA at Meta AI. These models apply LeCun’s self-supervised world-model approach to images and video.

Research Scientist & Manager, Meta (FAIR)

PhD · Ecole Normale Superieure (ENS) Paris — WILLOW team; advisors Ivan Laptev, Jean Ponce, Cordelia Schmid, Josef Sivic

FAIR Paris research lead. Coauthored fastText, DINO/DINOv2, and I-JEPA self-supervised models.

Research Scientist, Meta Superintelligence Labs

PhD · UC Berkeley — Alyosha Efros (BAIR)

Co-created OpenAI’s Sora and invented InstructPix2Pix. Now builds multimodal models and world models at Meta.

Co-led OpenAI’s Sora and now works on Google DeepMind’s Veo. These systems advance large-scale video generation as a world simulator.

Co-founder & CEO, Elorian

PhD · University of Edinburgh

Co-founder and CEO of Elorian, which develops visual reasoning and world models. Previously led the Gemini data area at DeepMind.

Professor (EECS), UC Berkeley

PhD · MIT — MIT, advised by Eric Grimson

UC Berkeley EECS professor and BAIR co-founder and co-lead. A leading researcher in computer vision and deep learning.

Research Scientist, Meta

PhD · Stanford University — Fei-Fei Li

Stanford PhD who studied video generation (W.A.L.T) and embodied AI. Now develops world and simulation models at Meta.

PhD Candidate (final-year), UC San Diego

Created TD-MPC and TD-MPC2, scalable world models for continuous-control RL.

Created TD-MPC2 and Newt, scalable model-based RL world models for continuous control across hundreds of tasks.

Research Scientist, Google DeepMind · Assistant Professor, Columbia University

PhD · University of Washington — GRAIL / UW Graphics (advisors: Steve Seitz, Brian Curless, Rick Szeliski)

Google DeepMind research scientist and Columbia professor. Studies generative video, 3D models, and world models.

Researcher, General Intuition

PhD · University of Cambridge — Machine Intelligence Lab (advisor: Roberto Cipolla)

Led GAIA generative world models for autonomous driving at Wayve. Now builds world models at General Intuition.

Founder & CEO, Morpheus

PhD · Cornell University — advised by Serge Belongie

Created AdaIN, now used across diffusion models. Founded Morpheus, which develops interactive video world models that run in real time.

Associate Professor of EECS, MIT

PhD · MIT — Ted Adelson (Brain & Cognitive Sciences)

Co-created pix2pix and CycleGAN. Studies shared world representations through the Platonic Representation Hypothesis.

Co-Founder, World Labs · Adjunct Assistant Professor, University of Michigan

PhD · Stanford University — Stanford Vision Lab (advisor: Fei-Fei Li)

Co-founded World Labs. Pioneered neural style transfer in real time and visual reasoning.

Co-founder of World Labs and a leader in 3D generation and visual reasoning. He coauthored widely used vision and neural-style-transfer methods.

Member of Technical Staff (Research Scientist), OpenAI

PhD · RWTH Aachen University — Mathematical Foundations of Computer Science (advisor: Erich Grädel)

Coauthored Attention Is All You Need, which introduced the Transformer. Research lead on OpenAI’s o1 reasoning models.

Co-founder & CEO, Wayve

PhD · University of Cambridge — Machine Intelligence Laboratory, advised by Roberto Cipolla

Co-founder and CEO of Wayve. Pioneered end-to-end deep learning for autonomous driving, called AV2.0.

Wayve CEO who designed GAIA, generative driving world models for autonomous vehicles.

Co-founder, World Labs

PhD · University of Tübingen — Max Planck Institute for Intelligent Systems / Bernstein Center for Computational Neuroscience, Tübingen

Co-founded World Labs. Studies computer vision and graphics, including 3D/4D reconstruction and neural rendering (Pulsar).

Assistant Professor of Computer Science, Columbia University

PhD · MIT — Advised by Antonio Torralba and Russ Tedrake (MIT CSAIL)

Develops structured and particle-based world models for robot manipulation of deformable objects and fluids.

Research Scientist, Google DeepMind

PhD · UC Berkeley — Pieter Abbeel

Created RingAttention and Large World Model (LWM), a multimodal video-language world model with a one-million-token context.

Co-Founder, World Labs

PhD · UC Berkeley — Berkeley graphics/imaging (advisor: Ren Ng)

Lead author of NeRF (Neural Radiance Fields). Co-founded World Labs.

Invented Neural Radiance Fields (NeRF), which launched modern neural 3D reconstruction. He co-founded World Labs.

Director, Research Scientist, Meta Superintelligence Labs (TBD Labs)

PhD · Carnegie Mellon University — Robotics Institute (advised by Martial Hebert / Abhinav Gupta)

Studies self-supervised vision. Coauthored I-JEPA, DINO, and Movie Gen / Emu Video generation.

Research Scientist, Open-Endedness Team, Google DeepMind · Honorary Lecturer, University College London

PhD · University of Oxford — Stephen Roberts (Machine Learning Research Group)

Co-leads DeepMind’s Genie world models (Genie 1, 2, and 3) for generative interactive environments.

Leads Genie work at Google DeepMind. He advances interactive foundation world models that generate controllable environments.

Member of Technical Staff, Project Prometheus

PhD · Queen Mary University of London — Theoretical physics (string theory) PhD; later RL self-taught

Coauthored the DreamerV3 world-model RL agent and Memory Maze benchmark for long-term memory.

Former Head of Sora (departed April 2026), OpenAI

PhD · UC Berkeley — Alyosha Efros (BAIR)

Invented the Diffusion Transformer (DiT). Co-creator and head of OpenAI’s Sora.

Co-created the Diffusion Transformer (DiT) architecture, which underpins Sora and most modern video-generation models. He now works at OpenAI.

VP of World Models, AMI Labs

PhD · University of Wisconsin-Madison — Robert Nowak (Electrical Engineering)

Co-created I-JEPA and V-JEPA self-supervised world models. Now VP World Models at LeCun’s AMI Labs.

Director of Research, Creative Vision, Snap Research

PhD · University of Trento

Director of Research at Snap. Pioneered video generation through MoCoGAN, First Order Motion Model, and Snap Video.

Co-Founder & CEO, Runway

Co-founder and CEO of Runway. Extends generative video into general world models, including GWM-1.

Runway CEO who ships commercial generative world models for video, including the Gen-series and GWM-1. They are widely used in film and media production.

Research Scientist, Meta (FAIR) · Adjunct Professor (Mila founding member), Universite de Montreal / Mila

PhD · Universite de Montreal — Yoshua Bengio (Mila)

Pioneered denoising autoencoders and denoising score matching. Coauthored I-JEPA self-supervised models.

Assistant Professor of Computer Science (courtesy in Psychology), Stanford University

PhD · MIT — William T. Freeman and Joshua B. Tenenbaum

Studies physical scene understanding and visual reasoning with physics engines and neuro-symbolic methods.

Leads physics-based and 3D world-model research at Stanford. This includes WonderZoom for multi-scale 3D world generation and models of intuitive physics.

Co-founder & CEO, Genesis AI

PhD · Carnegie Mellon University — Robotics Institute, advised by Katerina Fragkiadaki

Co-founder and CEO of Genesis AI. Created RoboGen and the Genesis generative physics engine for robotics.

Co-founder & Chief Science Officer, AMI Labs · Assistant Professor of Computer Science (on leave), New York University

PhD · UC San Diego — Zhuowen Tu

Co-created Diffusion Transformers (DiT) and ConvNeXt. Co-founder and CSO of AMI Labs, which develops JEPA world models.

Co-created V-JEPA and leads work on generative representation learning. Recent flow-map distillation and representation autoencoders set new state-of-the-art results for image generation.

Assistant Professor of Computer Science, NYU Courant · Staff Research Scientist, Google DeepMind

PhD · UC Berkeley — Pieter Abbeel

Develops generative world models and real-world simulators, including UniSim, for embodied agents.

Pioneered the use of video generation as a universal world model and simulator. She coauthored Gemini Robotics and Veo-based policy evaluation at Google DeepMind.

Co-Founder, General Intuition

PhD · University of Geneva — François Fleuret's group

Led DIAMOND and IRIS diffusion and transformer world models. Co-founded General Intuition.

First author of DIAMOND (NeurIPS 2024), a diffusion world model that achieves human-level Atari play.

Incoming Assistant Professor, Imperial College London · Research Scientist, Meta FAIR

PhD · Tel Aviv University / UC Berkeley — Advised by Amir Globerson and Trevor Darrell

Contributed Navigation World Models, which received a CVPR 2025 Best Paper Honorable Mention. Studies self-supervised visual world modeling.

First author of Navigation World Models (CVPR 2025 Honorable Mention). It is a billion-parameter conditional diffusion transformer for visual navigation prediction.

Dickinson Family Assistant Professor, Duke University

PhD · Columbia University — Hod Lipson's Creative Machines Lab

Dickinson Family Assistant Professor at Duke and General Robotics Lab director. Studies self-modeling robots and automated scientific discovery.

First author of Diffusion Forcing (NeurIPS 2024), a training approach that combines video prediction and world models. Recent work includes large video planners for robot control.

Assistant Professor, Harvard University · Senior Research Scientist, Google DeepMind

PhD · Massachusetts Institute of Technology — Leslie Kaelbling, Tomas Lozano-Perez & Joshua Tenenbaum (MIT EECS)

Harvard SEAS Assistant Professor and Kempner Investigator. Develops compositional and energy-based generative world models.

Pioneered composable diffusion and energy-based world models. He frequently collaborates on robotics work that uses video models as simulators.

Research Engineer, Google DeepMind

DeepMind researcher and technical lead or contributor on Genie 2 and Genie 3 foundation world models.

Known for Genie 2 (2024). Technical lead on Genie interactive world models.

Co-Founder, Genesis AI · PhD Candidate, MIT CSAIL

PhD · Massachusetts Institute of Technology — Daniela Rus (MIT CSAIL Distributed Robotics Lab)

MIT CSAIL roboticist and co-founder of Genesis AI, a startup that develops universal robotics foundation models.

Co-leads the Genesis physics engine for embodied AI. Reports indicate it runs 10–80 times faster than prior simulators.

Video & Inference Lead, HPC-AI Tech

Founded and leads Open-Sora, the first open-source video-generation model similar to Sora.

Known for Open-Sora (2024). Leads Open-Sora, an open-source video-generation framework.

PhD Student, New York University

Lead author of DINO-WM world models for zero-shot planning with pretrained visual features.

Known for Navigation World Models (CVPR 2025). Coauthored navigation world models with Amir Bar.