World Model (Artificial Intelligence)

World models are machine learning systems that build internal representations of environments to predict changes over time, enabling agents to plan and act without real-world trial and error. Jürgen Schmidhuber introduced the term in 1990, and David Ha and Schmidhuber revived it in a 2018 paper. Google DeepMind's Genie 3, introduced in August 2025, generates photorealistic, real-time interactive worlds from text prompts at 24 frames per second, and Waymo adopted it in February 2026 to create the Waymo World Model for autonomous driving simulation.

World model artificial intelligence A world model in artificial intelligence https://en.wikipedia.org/wiki/Artificial intelligence is a machine learning https://en.wikipedia.org/wiki/Machine learning system that builds an internal representation of an environment. Often this is via understanding objects within video, which predictive LLMs cannot. The model predicts how that environment changes over time in response to actions. Researchers design world models to help agents plan, reason, and act without constant real-world trial and error. World models differ from systems that merely classify or generate outputs. They simulate dynamics such as physics, object interactions, and causality. Early ideas date to the 1990s. Modern versions power robots, autonomous driving, and interactive video generation. History edit </w/index.php?title=World model artificial intelligence &action=edit§ion=1 Jürgen Schmidhuber https://en.wikipedia.org/wiki/Jürgen Schmidhuber introduced the term world model in machine learning https://en.wikipedia.org/wiki/Machine learning in 1990.<sup \ 1\ cite note-1 </sup He proposed recurrent neural networks https://en.wikipedia.org/wiki/Recurrent neural networks that predict future states from observations and use those predictions to train agents. David Ha and Schmidhuber revived the concept in a 2018 paper. Their agents learned to drive virtual cars and play video games inside self-generated simulations. \ 2\ cite note-ha-schmidhuber-2018-2 Yann LeCun https://en.wikipedia.org/wiki/Yann LeCun advanced the idea in a 2022 position paper titled "A Path Towards Autonomous Machine Intelligence".<sup \ 3\ cite note-:1-3 </sup He argued that intelligence requires predictive models of the world rather than pure pattern matching https://en.wikipedia.org/wiki/Pattern matching . LeCun proposed the joint embedding predictive architecture JEPA as a practical foundation. LeCun and collaborators developed several JEPA variants. V-JEPA 2 reached state-of-the-art performance on video understanding and physical reasoning at the time.<sup \ 4\ cite note-:0-4 </sup It supports zero-shot robot control in unfamiliar environments.<sup \ 4\ cite note-:0-4 </sup Introduced in March 2026, LeWorldModel trains stably end-to-end from raw pixels and uses two loss terms and avoids hand-crafted heuristics.<sup \ 5\ cite note-5 </sup LeCun founded Advanced Machine Intelligence Labs https://en.wikipedia.org/wiki/Advanced Machine Intelligence Labs in 2026 to further develop world models. \ 6\ cite note-6 \ 7\ cite note-7 Google DeepMind https://en.wikipedia.org/wiki/Google DeepMind introduced Genie <https://en.wikipedia.org/wiki/Genie AI model in 2024. The model learned interactive environments from unlabeled internet videos. Genie 2 followed in late 2024 and added three-dimensional generation. The Genie series set benchmarks for general-purpose simulation. Genie 3 was introduced in August 2025. It produces photorealistic, real-time interactive worlds from text prompts which are displayed at 24 frames per second and explored in real time with text or image prompts. The model supports persistent three-dimensional worlds and real-time interaction. \ 8\ cite note-at-3-8 Waymo https://en.wikipedia.org/wiki/Waymo adopted Genie 3 in February 2026 and used it to create a specialized world model for autonomous driving https://en.wikipedia.org/wiki/Autonomous driving simulation, called the Waymo World Model https://en.wikipedia.org/wiki/Waymo World Model . It produces synchronized camera and lidar https://en.wikipedia.org/wiki/Lidar outputs and creates edge cases that real robotaxis https://en.wikipedia.org/wiki/Robotaxi rarely encounter. The edge cases were reported to be unusual by PCMag https://en.wikipedia.org/wiki/PCMag . \ 9\ cite note-pcmag-wymo-9 General Intuition https://en.wikipedia.org/wiki/General Intuition?action=edit&redlink=1 announced a $133.7 million seed round.<sup \ 10\ cite note-tc-gi-10 </sup World Labs raised $1 billion. AMI raised $1.03 billion. \ 11\ cite note-11 In April 2026, Alibaba https://en.wikipedia.org/wiki/Alibaba Group announced Happy Oyster, its world model designed for real-time and “flowy” world model. It includes a directing mode for world building based on text and image prompts and a wandering mode for exploring the resulting world. It can generate 3-minute in-world video clips. \ 12\ cite note-:2-12 Also in April, World Labs, co-founded by Li Fei Fei https://en.wikipedia.org/wiki/Fei-Fei Li , unveiled Spark 2.0, an open-source 3D Gaussian splatting https://en.wikipedia.org/wiki/Gaussian splatting rendering engine https://en.wikipedia.org/wiki/Rendering engine that targets smartphone-class devices. \ 12\ cite note-:2-12 In June 2026, Nvidia https://en.wikipedia.org/wiki/Nvidia released Cosmos 3, a family of open-weight models. It combines previously independent physical reasoning, world simulation, and action generation. Cosmos 3 integrates can process and generate text, image, video, audio, and action sequences. The model employs a Mixture-of-Transformers" MoT approach. An autoregressive AR transformer https://en.wikipedia.org/wiki/Transformer handles reasoning and next-token prediction, while a diffusion https://en.wikipedia.org/wiki/Diffusion model transformer DT does multimodal generation. Encoders ViT https://en.wikipedia.org/wiki/Vision transformer for vision, VAE https://en.wikipedia.org/wiki/Variational autoencoder for visual/audio, and domain-specific for actions and generate a shared representation space using 3D multi-dimensional rotary position embedding mRoPE for spatial and temporal information. The family includes Cosmos3-Nano 16B parameters for workstations; Cosmos3-Super 64B parameters for research. \ 13\ cite note-13 Conceptualization edit </w/index.php?title=World model artificial intelligence &action=edit§ion=2 World models broadly refer to internal representations used to predict or simulate an environment, but the exact definition remains unsettled and continues to evolve as understeanding grows across artificial intelligence, robotics, cognitive science and computational neuroscience. \ 14\ cite note-14 \ 15\ cite note-:6-15 <sup \ 16\ cite note-:3-16 </sup World-model learning and inference were used broadly to study how systems predicts states, interpret sensory information and guide action. \ 16\ cite note-:3-16 \ 17\ cite note-:4-17 <sup \ 18\ cite note-:5-18 </sup Predictive coding, where predictions are checked against observations, and differences betweeen them used to update internal moels, provides one framework for comparing world-model learning across artificial intelligence, robotics and neuroscience. \ 16\ cite note-:3-16 \ 17\ cite note-:4-17 \ 18\ cite note-:5-18 In machine learning, a world model can be learned from experience and used to simulate prossible outcomes, allowing an agent to plan or learn behavior without testing every possibility directly in the environment. \ 19\ cite note-19 \ 2\ cite note-ha-schmidhuber-2018-2 \ 20\ cite note-20 <sup \ 21\ cite note-21 </sup World modeling therefore describes a general approach to representation, learning and decision-making rather than a particular model architecture. \ 15\ cite note-:6-15 <sup \ 3\ cite note-:1-3 </sup Additional examples of this approach include planning and behavior learning within learned latent models. \ 22\ cite note-22 \ 23\ cite note-23 Architecture edit </w/index.php?title=World model artificial intelligence &action=edit§ion=3 World models process raw sensory data such as video frames or lidar scans. They compress this input into compact latent representations. The system then predicts future representations rather than pixel-by-pixel reconstructions. Many modern world models use joint embedding predictive architecture JEPA . An encoder turns observations into embeddings . A predictor estimates one or a suite of embeddings from the current one and an action. In some cases a critic chooses one embedding as the best result. A regularizer keeps embeddings well-behaved. \ 4\ cite note-:0-4 The model trains by minimizing prediction error in embedding space. This approach avoids the high cost of generating every detail. Some architectures add explicit components. A fast reactive path handles immediate responses. A slower deliberative path performs longer-horizon planning. Video prediction accuracy or robot success rates are key metrics, but do not always predict real-world performance. Generative world models such as Genie 3 combine these with a simulator . They accept text prompts or layouts and output consistent video, lidar, or three-dimensional scenes. World models often train with self-supervised learning. They use large unlabeled datasets of video or robot interactions. Self-supervised learning https://en.wikipedia.org/wiki/Self-supervised learning can speed learning. Reinforcement learning https://en.wikipedia.org/wiki/Reinforcement learning can fine-tune a model for specific tasks. Applications edit </w/index.php?title=World model artificial intelligence &action=edit§ion=4 World models support robot learning. Agents train inside simulations and transfer skills to the physical world. This reduces the need for dangerous or expensive real-world trials. Autonomous vehicles use world models to test rare events.<sup \ 24\ cite note-24 </sup Waymo's system simulates tornadoes or unusual pedestrian behavior. Companies train planners without putting vehicles on public roads. Interactive entertainment benefits from world models. Genie 3 lets users generate playable environments from simple descriptions. Game studios prototype levels faster. Scientific simulation gains from these models. Researchers model physical systems or biological processes at scale. Planners in logistics or urban design test strategies inside accurate digital twins. Comparison with large language models edit </w/index.php?title=World model artificial intelligence &action=edit§ion=5 Both world models and large language models https://en.wikipedia.org/wiki/Large language models LLMs use inferencing on their inputs to make predictions. LLMs operate on textual inputs. They predict the next token in text sequences. They excel at language-oriented tasks such as translation or summarization. However, they lack understanding of physics. World models operate on sensor inputs such as pixels. They predict state changes in that data in latent space. This design supports planning and causal reasoning https://en.wikipedia.org/wiki/Causal reasoning . LLMs generate fluent text but often fail at consistent physical predictions. Their architecture employs transformers <https://en.wikipedia.org/wiki/Transformer deep learning with refinements such as mixture of experts https://en.wikipedia.org/wiki/Mixture of experts . World models divide an inferencing task into work performed by encoders, predictors, simulators, and other pieces. They typically handle multimodal inputs such as video, lidar, radar, and audio, guided by textual prompting. LLMs power chatbots https://en.wikipedia.org/wiki/Chatbots and code assistants. World models drive embodied agents that act in dynamic environments, such as autonomous driving. The two may be combined in hybrid systems. For example, a LLM handles instructions, while a world model manages low-level control. World model proponents such as LeCun claim that because LLMs are trained only on text, they have no ability to predict anything beyond text, such as real-world events. \ 3\ cite note-:1-3 \ 25\ cite note-newsweek-lecun-25 Benchmarks edit </w/index.php?title=World model artificial intelligence &action=edit§ion=6 World model benchmarks test physical understanding, long-term consistency, planning, and generalization from sensor data. Meta introduced three benchmarks for V-JEPA 2. \ 26\ cite note-26 - IntPhys 2 measures a model's ability to detect physics violations. It presents pairs of videos that diverge when one breaks physical rules. Humans score near 100% accuracy. V-JEPA 2 achieves little better than random chance on many conditions.<sup \ 27\ cite note-27 </sup - Minimal Video Pairs MVPBench tests physical understanding through multiple-choice questions based on short video clips. It probes object interactions and causality.<sup \ 28\ cite note-28 </sup - Something-Something tests action recognition.<sup \ 29\ cite note-29 </sup - Epic-Kitchens-100 tests human action anticipation. DeepMind benchmark: - Interactive evaluation measures consistency over minutes of interaction, memory of off-screen objects, and response to user actions or text prompts.<sup \ 30\ cite note-30 </sup Waymo benchmark: - Output generation quality: Metrics include realism, controllability via text prompts , and usefulness for training planners in simulated worlds. However, pixel reconstruction error rate with episodic rewards often fails. Other: - Epic-Kitchens-100 often measured with Recall@5 <sup \ 31\ cite note-31 </sup - Ego4D - 50 Salads, Breakfast, etc. Potential benchmarks: - Zero-shot transfer to robots - Long-horizon planning - Implausible prediction rate See also edit </w/index.php?title=World model artificial intelligence &action=edit§ion=7 References edit </w/index.php?title=World model artificial intelligence &action=edit§ion=8 1. ↑ cite ref-1 "1990: Planning & Reinforcement Learning with Recurrent World Models and Artificial Curiosity" https://web.archive.org/web/20260308181935/https://people.idsia.ch/~juergen/world-models-planning-curiosity-fki-1990.html . people.idsia.ch . Archived from the original https://people.idsia.ch/~juergen/world-models-planning-curiosity-fki-1990.html on 2026-03-08. Retrieved 2026-03-25. 2. 1 cite ref-ha-schmidhuber-2018 2-0 2 cite ref-ha-schmidhuber-2018 2-1 Ha, David; Schmidhuber, Jürgen 2018 . "Recurrent World Models Facilitate Policy Evolution" https://proceedings.neurips.cc/paper/2018/hash/2de5d16682c3c35007e4e92982f1a2ba-Abstract.html . Advances in Neural Information Processing Systems 31 NeurIPS 2018 . Curran Associates. pp. 2451–2463. arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 1809.01999 https://arxiv.org/abs/1809.01999 . 3. 1 cite ref-:1 3-0 2 cite ref-:1 3-1 3 cite ref-:1 3-2 LeCun, Yann June 27, 2022 . "A Path Towards Autonomous Machine Intelligence Version 0.9.2" https://awxlong.github.io/assets/pdf/autonomous-ml-lecunnpdf.pdf PDF . 4. 1 cite ref-:0 4-0 2 cite ref-:0 4-1 3 cite ref-:0 4-2 "Introducing the V-JEPA 2 world model and new benchmarks for physical reasoning" https://ai.meta.com/blog/v-jepa-2-world-model-benchmarks/ . Meta AI. June 11, 2025. Retrieved March 25, 2026. 5. ↑ cite ref-5 Maes, Lucas; Lidec, Quentin Le; Scieur, Damien; LeCun, Yann; Balestriero, Randall 2026-03-13 . "LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels". arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 2603.19312 https://arxiv.org/abs/2603.19312 cs.LG https://arxiv.org/archive/cs.LG . 6. ↑ cite ref-6 Metz, Cade March 10, 2026 . "Former Meta A.I. Chief's Start-Up Is Valued at $3.5 Billion" https://www.nytimes.com/2026/03/10/technology/ami-labs-yann-lecun-funding.html . The New York Times – via NYTimes.com. 7. ↑ cite ref-7 Orru, Mauro March 10, 2026 . "Former Meta AI Pioneer Yann LeCun Raises Over $1 Billion for New Startup" https://www.wsj.com/tech/ai/yann-lecuns-ai-startup-raises-over-1-billion-14928928 . The Wall Street Journal . 8. ↑ cite ref-at-3 8-0 Whitwam, Ryan 2025-08-05 . "DeepMind reveals Genie 3 "world model" that creates real-time interactive simulations" https://arstechnica.com/ai/2025/08/deepmind-reveals-genie-3-world-model-that-creates-real-time-interactive-simulations/ . Ars Technica . 9. ↑ cite ref-pcmag-wymo 9-0 Martindale, Jon February 6, 2026 . "Waymo Is Using Google's Genie 3 AI to Practice Handling Tornadoes, Elephants" https://www.pcmag.com/news/waymo-is-using-googles-genie-3-ai-to-practice-handling-tornadoes-elephants . PCMag . Retrieved April 6, 2026. 10. ↑ cite ref-tc-gi 10-0 Bellan, Rebecca October 16, 2025 . "General Intuition lands $134M seed to teach agents spatial reasoning using video game clips" https://techcrunch.com/2025/10/16/general-intuition-lands-134m-seed-to-teach-agents-spatial-reasoning-using-video-game-clips/ . TechCrunch . Retrieved June 30, 2026. 11. ↑ cite ref-11 McCormick, Packy 19 March 2026 . "World Models: Computing the Uncomputable" https://www.notboring.co/p/world-models . www.notboring.co . Retrieved 2026-04-09. 12. 1 cite ref-:2 12-0 2 cite ref-:2 12-1 "Chinese tech giants, AI 'godmother' Li Fei-Fei race into world models" https://www.scmp.com/tech/big-tech/article/3350351/chinese-tech-giants-ai-godmother-li-fei-fei-race-seize-edge-world-models . South China Morning Post . 2026-04-16. Retrieved 2026-04-18. 13. ↑ cite ref-13 "Welcome NVIDIA Cosmos 3: The First Open Omni-model for Physical AI Reasoning and Action" https://huggingface.co/blog/nvidia/cosmos-3-for-physical-ai . huggingface.co . 2026-06-01. Retrieved 2026-06-01. 14. ↑ cite ref-14 Sakagami, Ryo; Lay, Florian S.; Dömel, Andreas; Schuster, Martin J.; Albu-Schäffer, Alin; Stulp, Freek 2023-11-02 . "Robotic world models—conceptualization, review, and engineering best practices" https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10652279 . Frontiers in Robotics and AI . 10 1253049. Frontiers. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.3389/frobt.2023.1253049 https://doi.org/10.3389%2Ffrobt.2023.1253049 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 2296-9144 https://search.worldcat.org/issn/2296-9144 . PMC <https://en.wikipedia.org/wiki/PMC identifier 10652279 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10652279 . PMID <https://en.wikipedia.org/wiki/PMID identifier 38023585 https://pubmed.ncbi.nlm.nih.gov/38023585 . 15. 1 cite ref-:6 15-0 2 cite ref-:6 15-1 Moerland, Thomas M.; Broekens, Joost; Plaat, Aske; Jonker, Catholijn M. 2023-01-04 . "Model-based Reinforcement Learning: A Survey" https://doi.org/10.1561/2200000086 . Foundations and Trends in Machine Learning . 16 1 : 1–118. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1561/2200000086 https://doi.org/10.1561%2F2200000086 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 1935-8237 https://search.worldcat.org/issn/1935-8237 . 16. 1 cite ref-:3 16-0 2 cite ref-:3 16-1 3 cite ref-:3 16-2 Taniguchi, Tadahiro; Murata, Shingo; Suzuki, Masahiro; Ognibene, Dimitri; Lanillos, Pablo; Ugur, Emre; Jamone, Lorenzo; Nakamura, Tomoaki; Ciria, Alejandra; Lara, Bruno; Pezzulo, Giovanni 2023-07-03 . "World models and predictive coding for cognitive and developmental robotics: frontiers and challenges" https://www.tandfonline.com/doi/full/10.1080/01691864.2023.2225232 . Advanced Robotics . 37 13 : 780–806. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1080/01691864.2023.2225232 https://doi.org/10.1080%2F01691864.2023.2225232 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 0169-1864 https://search.worldcat.org/issn/0169-1864 . 17. 1 cite ref-:4 17-0 2 cite ref-:4 17-1 Friston, Karl; Moran, Rosalyn J.; Nagai, Yukie; Taniguchi, Tadahiro; Gomi, Hiroaki; Tenenbaum, Josh 2021-12-01 . "World model learning and inference" https://www.sciencedirect.com/science/article/pii/S0893608021003610 . Neural Networks . 144 : 573–590. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1016/j.neunet.2021.09.011 https://doi.org/10.1016%2Fj.neunet.2021.09.011 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 0893-6080 https://search.worldcat.org/issn/0893-6080 . PMID <https://en.wikipedia.org/wiki/PMID identifier 34634605 https://pubmed.ncbi.nlm.nih.gov/34634605 . 18. 1 cite ref-:5 18-0 2 cite ref-:5 18-1 Ohmae, Shogo; Ohmae, Keiko August 2026 . "Brain-AI convergence: Generative world models and hierarchical attention for human intelligence" https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13494634 . Patterns . 7 8 101593. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1016/j.patter.2026.101593 https://doi.org/10.1016%2Fj.patter.2026.101593 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 2666-3899 https://search.worldcat.org/issn/2666-3899 . PMC <https://en.wikipedia.org/wiki/PMC identifier 13494634 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13494634 . PMID <https://en.wikipedia.org/wiki/PMID identifier 42630565 https://pubmed.ncbi.nlm.nih.gov/42630565 . 19. ↑ cite ref-19 Sutton, Richard S. 1990 . "Integrated Architectures for Learning, Planning, and Reacting Based on Approximating Dynamic Programming" https://mlanthology.org/icml/1990/sutton1990icml-integrated/ . Machine Learning Proceedings 1990 . pp. 216–224. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1016/B978-1-55860-141-3.50030-4 https://doi.org/10.1016%2FB978-1-55860-141-3.50030-4 . ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-1-55860-141-3 https://en.wikipedia.org/wiki/Special:BookSources/978-1-55860-141-3 . 20. ↑ cite ref-20 Schrittwieser, Julian; Antonoglou, Ioannis; Hubert, Thomas; Simonyan, Karen; Sifre, Laurent; Schmitt, Simon; Guez, Arthur; Lockhart, Edward; Hassabis, Demis; Graepel, Thore; Lillicrap, Timothy; Silver, David December 2020 . "Mastering Atari, Go, chess and shogi by planning with a learned model" https://www.nature.com/articles/s41586-020-03051-4 . Nature . 588 7839 . Nature Publishing Group: 604–609. arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 1911.08265 https://arxiv.org/abs/1911.08265 . Bibcode <https://en.wikipedia.org/wiki/Bibcode identifier : 2020Natur.588..604S https://ui.adsabs.harvard.edu/abs/2020Natur.588..604S . doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1038/s41586-020-03051-4 https://doi.org/10.1038%2Fs41586-020-03051-4 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 1476-4687 https://search.worldcat.org/issn/1476-4687 . PMID <https://en.wikipedia.org/wiki/PMID identifier 33361790 https://pubmed.ncbi.nlm.nih.gov/33361790 . 21. ↑ cite ref-21 Hafner, Danijar; Pasukonis, Jurgis; Ba, Jimmy; Lillicrap, Timothy April 2025 . "Mastering diverse control tasks through world models" https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12003158 . Nature . 640 8059 . Nature Publishing Group: 647–653. Bibcode <https://en.wikipedia.org/wiki/Bibcode identifier : 2025Natur.640..647H https://ui.adsabs.harvard.edu/abs/2025Natur.640..647H . doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1038/s41586-025-08744-2 https://doi.org/10.1038%2Fs41586-025-08744-2 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 1476-4687 https://search.worldcat.org/issn/1476-4687 . PMC <https://en.wikipedia.org/wiki/PMC identifier 12003158 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12003158 . PMID <https://en.wikipedia.org/wiki/PMID identifier 40175544 https://pubmed.ncbi.nlm.nih.gov/40175544 . 22. ↑ cite ref-22 Hafner, Danijar; Lillicrap, Timothy; Fischer, Ian; Villegas, Ruben; Ha, David; Lee, Honglak; Davidson, James 2019-05-24 . "Learning Latent Dynamics for Planning from Pixels" https://proceedings.mlr.press/v97/hafner19a.html . Proceedings of the 36th International Conference on Machine Learning . PMLR: 2555–2565. ISSN <https://en.wikipedia.org/wiki/ISSN identifier 2640-3498 https://search.worldcat.org/issn/2640-3498 . 23. ↑ cite ref-23 Hafner, Danijar; Lillicrap, Timothy; Fischer, Ian; Villegas, Ruben; Ha, David; Lee, Honglak; Davidson, James 2019-05-24 . "Learning Latent Dynamics for Planning from Pixels" https://proceedings.mlr.press/v97/hafner19a.html . Proceedings of the 36th International Conference on Machine Learning . PMLR: 2555–2565. ISSN <https://en.wikipedia.org/wiki/ISSN identifier 2640-3498 https://search.worldcat.org/issn/2640-3498 . 24. ↑ cite ref-24 Yin, Tenny; Mei, Zhiting; Zheng, Zhonghe; Yamane, Miyu; Wang, David; Sceats, Jade; Bateman, Samuel M.; Zha, Lihan; Badithela, Apurva; Shorinwa, Ola; Majumdar, Anirudha. "PlayWorld: Learning Robot World Models from Autonomous Play" https://arxiv.org/html/2603.09030v1 . arxiv.org . Retrieved 2026-03-25. 25. ↑ cite ref-newsweek-lecun 25-0 Weldon, Marcus April 2, 2025 . "AI 'Godfather' Yann LeCun: LLMs Are Nearing the End, but Better AI Is Coming" https://www.newsweek.com/nw-ai/ai-impact-interview-yann-lecun-llm-limitations-analysis-2054255 . Newsweek . Retrieved June 30, 2026. 26. ↑ cite ref-26 "Introducing the V-JEPA 2 world model and new benchmarks for physical reasoning" https://ai.meta.com/blog/v-jepa-2-world-model-benchmarks/ . Meta AI. June 11, 2025. Retrieved March 26, 2026. 27. ↑ cite ref-27 Bordes, Florian; Garrido, Quentin; Kao, Justine T.; Williams, Adina; Rabbat, Michael; Dupoux, Emmanuel 2025 , IntPhys 2: Benchmarking Intuitive Physics Understanding in Complex Synthetic Environments , arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 2506.09849 https://arxiv.org/abs/2506.09849 28. ↑ cite ref-28 Krojer, Ben 2025 . "A Shortcut-aware Video-QA Benchmark for Physical Understanding via Minimal Video Pairs". arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 2506.09987 https://arxiv.org/abs/2506.09987 cs.CV https://arxiv.org/archive/cs.CV . 29. ↑ cite ref-29 Goyal, Raghav 2017 . "The "something something" video database for learning and evaluating visual common sense". Proceedings of the IEEE International Conference on Computer Vision ICCV . arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 1706.04261 https://arxiv.org/abs/1706.04261 . 30. ↑ cite ref-30 "Genie 3: A new frontier for world models" https://deepmind.google/blog/genie-3-a-new-frontier-for-world-models/ . Google DeepMind. August 5, 2025. Retrieved March 26, 2026. 31. ↑ cite ref-31 "EPIC-KITCHENS-100 Dataset" https://epic-kitchens.github.io/ . Retrieved March 26, 2026. External links edit </w/index.php?title=World model artificial intelligence &action=edit§ion=9 - Carmack, John March 31, 2026 . "Paper review: LeWorldModel: Stable End-to-End joined-Embedding Predictive Architecture from Pixels" https://x.com/ID AA Carmack/status/2039046172799578122?s=20 . X . Retrieved April 1, 2026.