---
id: AGI-WM-001
title: "World Models"
alternative_names:
  - "Learned World Models"
  - "Environment Models"
  - "Predictive World Models"
category: "Advanced Intelligence → General Intelligence"
volume: V
chapter: 27
entry_type: "Standard Article"
level: 3
status: "Draft"
last_reviewed: 2026-10-15
version: "0.1.0"
prerequisites:
  - "AGI-PATH-001"
  - "TECH-RL-001"
related_entries:
  - "AGI-REQ-001"
  - "AGI-CL-001"
  - "TECH-DIFF-001"
  - "TECH-ROBOT-001"
---

# World Models

## 1. Concise Definition
A **world model** is a learned or constructed internal model that predicts how the state of an environment changes over time and in response to an agent's actions, and that can be used to plan, imagine outcomes or interpret observations. In AI research the term covers explicit latent-dynamics models used in reinforcement learning, video- and interaction-generating models, and the question of whether large language models acquire implicit models of the processes that generated their training data.

## 2. Overview & Significance
Planning requires predicting consequences. An agent that can predict what will happen if it acts can evaluate options without executing them, which reduces the number of costly or dangerous real-world trials. Proposals for general machine intelligence frequently place a world model at the centre of the architecture (LeCun, 2022), and [AGI-PATH-001](../vol05_agi/AGI-PATH-001_pathways_to_agi.md) lists world-model approaches as one of several pathways. Whether a world model is *necessary* for general intelligence is Contested (see [AGI-REQ-001](../vol05_agi/AGI-REQ-001_requirements_for_general_intelligence.md)); this article describes what is meant by the term and what has been demonstrated.

Three related meanings are often conflated.

1. **Control-oriented world models.** A model of dynamics learned so that an agent can plan or train a policy in imagination (Ha & Schmidhuber, 2018; Hafner et al., 2023).
2. **Generative environment models.** Models that produce video or interactive environments conditioned on text, images or actions (Bruce et al., 2024).
3. **Implicit world models in sequence predictors.** The hypothesis that a model trained only to predict the next token or move develops internal structure that mirrors the underlying world (Li et al., 2023; Gurnee & Tegmark, 2023; Vafa et al., 2024).

## 3. Historical Development
The idea of an agent with an internal model of its environment is long-standing in cognitive science and control theory. In the neural-network literature, Ha and Schmidhuber (2018) presented a system in which a generative model of the environment is learned from observations and a small controller is trained within the model's own simulated dynamics. The same period saw model-based reinforcement learning advance in board games: MuZero learned a model sufficient for planning without being given the rules of the game, and reached strong performance on Go, chess, shogi and Atari (Schrittwieser et al., 2020).

Hafner et al. (2023) introduced DreamerV3, which uses a learned world model with a single fixed configuration across many tasks, and which the authors state is the first algorithm to collect diamonds in Minecraft from scratch without human data or curricula. LeCun (2022) proposed an architecture for autonomous agents built around a world model trained with joint-embedding predictive methods rather than pixel-level generation. Generative video and interactive-environment models followed: Genie is described by its authors as an approximately 11-billion-parameter foundation world model trained on unlabelled Internet video that can generate action-controllable virtual worlds from prompts (Bruce et al., 2024). V-JEPA 2 extended joint-embedding methods to video and, with a small amount of robot data, to planning in robot manipulation (Assran et al., 2025).

In parallel, interpretability studies asked whether sequence models already contain world models. Li et al. (2023) trained a GPT-style model on Othello move sequences and found internal board-state representations that could be intervened on. Gurnee and Tegmark (2023) reported linear representations of space and time in Llama-2 models. Vafa et al. (2024) proposed metrics based on the Myhill-Nerode theorem and found that models which do well on standard tests had much less coherent implicit world models than those tests suggested.

## 4. Technical / Conceptual Explanation

### 4.1 Formal structure
A world model is commonly formalised as a stochastic transition model over latent states. Given observations $o_t$, actions $a_t$ and latent state $z_t$, a typical model comprises:

- an encoder, $z_t \sim q_\phi(z_t \mid z_{t-1}, a_{t-1}, o_t)$;
- a dynamics (prior) model, $\hat{z}_t \sim p_\theta(\hat{z}_t \mid z_{t-1}, a_{t-1})$;
- a decoder or predictive head, $\hat{o}_t \sim p_\theta(\hat{o}_t \mid z_t)$, and a reward predictor $\hat{r}_t$.

Training maximises a variational objective that balances reconstruction or prediction accuracy against closeness between the posterior and prior over latent states. In this family the learned dynamics can be rolled forward from a starting state to produce imagined trajectories, and a policy is optimised on those trajectories (Ha & Schmidhuber, 2018; Hafner et al., 2023). This describes the general design; the specific objectives, architectures and normalisation tricks differ by system.

### 4.2 Planning with a learned model
Given a model, an agent chooses actions by search or optimisation over imagined futures. MuZero combines a learned model of reward, value and policy-relevant quantities with Monte Carlo tree search (Schrittwieser et al., 2020): the learned model need not reconstruct observations, only to predict what matters for planning. This shows that a world model, in the sense of a model useful for decisions, need not be a faithful simulator.

### 4.3 Prediction in representation space
LeCun (2022) argues for predicting in an abstract representation space instead of predicting every detail of the input, because many details of the world are not predictable. V-JEPA 2 follows this approach for video: the authors report pre-training on over 1 million hours of internet video and images, and a robot-planning variant post-trained on fewer than 62 hours of unlabelled robot video, deployed zero-shot on robot arms in two laboratories for pick-and-place tasks (Assran et al., 2025). These are results reported by the developing group and are Preliminary.

### 4.4 Implicit world models
For language and sequence models, the claim is that internal activations encode latent variables of the data-generating process. Evidence takes the form of probes (decoding the variable from activations) and interventions (changing the representation and observing changes in output). Li et al. (2023) found a nonlinear board representation in the Othello setting; Gurnee and Tegmark (2023) found linear encodings of geographic and temporal variables in Llama-2 models. Vafa et al. (2024) tested coherence on game playing, logic puzzles and navigation, and found that the implicit models were much less consistent than standard tests indicated, which can cause failures on closely related tasks. Probing results show that information is present and decodable; they do not by themselves show that the model *uses* a coherent model for inference.

## 5. Types, Components & Classifications

| Type | Purpose | Example | Evidence level |
| :--- | :--- | :--- | :--- |
| Latent-dynamics model for control | Plan or train in imagination | Ha & Schmidhuber (2018); Hafner et al. (2023) | Well Supported in simulated environments |
| Value-equivalent model with search | Plan without reconstructing observations | Schrittwieser et al. (2020) | Well Supported in games |
| Joint-embedding predictive model | Predict in representation space; plan | LeCun (2022); Assran et al. (2025) | Hypothetical (architecture proposal); Preliminary (video results) |
| Generative interactive environment | Produce action-controllable worlds | Bruce et al. (2024) | Preliminary |
| Implicit world model in a sequence model | Latent structure inside LLMs or sequence models | Li et al. (2023); Gurnee & Tegmark (2023) | Contested |

A further classification concerns **explicit versus implicit** models, **deterministic versus stochastic** dynamics, and **observation-space versus latent-space** prediction. The choice among these determines the computational cost of rollouts and the kinds of errors the model compounds.

## 6. Applications & Concrete Examples
- **Game playing and simulated control.** MuZero and DreamerV3 are demonstrations in simulated or game environments, not deployed products (Schrittwieser et al., 2020; Hafner et al., 2023).
- **Generative environments.** Genie generates playable worlds from prompts; the authors also describe using its learned latent action space to train agents to imitate behaviours from unseen videos (Bruce et al., 2024). This is a laboratory demonstration.
- **Robot planning.** V-JEPA 2-AC, per its developers, performed zero-shot pick-and-place on Franka arms in two laboratories without data collected from those robots (Assran et al., 2025). This is a laboratory result reported by the developers; it has not been shown here to extend to open-ended tasks. For the broader field see [TECH-ROBOT-001](../vol03_technical/TECH-ROBOT-001_robotics_and_embodied_ai.md).
- **Interpretability studies.** The Othello and Llama-2 results (Li et al., 2023; Gurnee & Tegmark, 2023) are used as evidence in debates about what language models represent. See [FOUND-014](../vol01_foundations/FOUND-014_consciousness_and_machine_understanding.md) for the related philosophical question.

## 7. Advantages, Limitations & Unresolved Questions
**Advantages.** Planning in a model can make learning more sample-efficient than purely model-free learning; a model supports counterfactual evaluation of actions; and a single model can serve multiple tasks. (The claim of sample efficiency is an expected benefit of the approach; the sources reviewed here demonstrate performance but do not compare data efficiency in every case.)

**Limitations.**
- *Compounding error.* Small prediction errors accumulate over long rollouts, limiting the planning horizon.
- *Stochasticity and unpredictability.* Many aspects of the world cannot be predicted in detail, which motivates prediction in latent space (LeCun, 2022) but leaves open how to choose what to predict.
- *Evaluation difficulty.* A model may look plausible while being physically or logically inconsistent. Vafa et al. (2024) show standard task accuracy can overstate the coherence of implicit models.
- *Generative realism is not understanding.* A model that produces visually convincing video need not have an accurate model of physics. The sources reviewed here describe generation, not physics-accurate prediction.
- *Scale and cost.* Training large video models is computationally expensive; see [ECON-COMP-001](../vol09_economy/ECON-COMP-001_compute_economics.md).

**Open questions.**
1. Is a learned world model necessary for general intelligence, or can language-based reasoning over text and tools substitute for it? Contested.
2. Do language models contain usable world models, or only decodable correlates? Contested.
3. How should world-model quality be measured so that tests catch inconsistency rather than reward surface fluency?
4. Can world models learned from passive video support planning in the real world at scale?

## 8. Current State of Knowledge
As of 2026-10-15 the sources reviewed support the following. Learned world models enable strong performance in simulated and game environments (Schrittwieser et al., 2020; Hafner et al., 2023), on the authors' reports. Generative interactive models and video-based predictive models exist as research systems (Bruce et al., 2024; Assran et al., 2025). Probing studies find structured internal representations in language and sequence models (Li et al., 2023; Gurnee & Tegmark, 2023), while a metric-based study finds substantial incoherence in implicit models (Vafa et al., 2024). The overall picture of what implicit world models LLMs have is Contested. Figures such as parameter counts, training-data sizes and benchmark scores in the papers above are given at the dates of those papers; for current systems consult the Living Reference collections (`MODEL-001`, `FACT-001`). The architectures and training data of closed commercial video and world-model systems are, in general, undisclosed, and this article makes no claims about them.

## 9. Future Trajectory & Strategic Outlook
This section is descriptive and registers no forecast.

- **Near term (0 to 3 years).** Expect continued work on video-based predictive models, interactive-environment generators and benchmarks for physical and causal consistency.
- **Medium term (3 to 10 years).** Whether world-model approaches provide a path to general capability, or are one component alongside language-based reasoning and tools, will depend on evidence that does not yet exist in the sources reviewed.
- **Long term (10+ years).** Speculative; connections to embodied systems and scientific simulation are discussed in [TECH-ROBOT-001](../vol03_technical/TECH-ROBOT-001_robotics_and_embodied_ai.md) and [APP-SCI-001](../vol07_applications/APP-SCI-001_ai_in_science.md).

## 10. Related Entries & Knowledge Graph
- **Foundational Concepts**:
  - [[AGI-PATH-001] Pathways to AGI](../vol05_agi/AGI-PATH-001_pathways_to_agi.md): World-model approaches among proposed routes to AGI.
  - [[TECH-RL-001] Reinforcement Learning](../vol03_technical/TECH-RL-001_reinforcement_learning.md): Model-based reinforcement learning.
  - [[AGI-REQ-001] Requirements for General Intelligence](../vol05_agi/AGI-REQ-001_requirements_for_general_intelligence.md): World models as a candidate requirement.
- **Related Methods & Technologies**:
  - [[TECH-DIFF-001] Diffusion Models](../vol03_technical/TECH-DIFF-001_diffusion_models.md): A common basis of generative video models.
  - [[TECH-ROBOT-001] Robotics and Embodied AI](../vol03_technical/TECH-ROBOT-001_robotics_and_embodied_ai.md): Use of world models for robot planning.
  - [[AGI-CL-001] Continual Learning and Persistent Memory](../vol05_agi/AGI-CL-001_continual_learning_and_persistent_memory.md): Keeping a world model up to date.
- **Advanced Implications & Future Trajectories**:
  - [[FOUND-014] Consciousness and Machine Understanding](../vol01_foundations/FOUND-014_consciousness_and_machine_understanding.md): Whether internal models amount to understanding.
  - [[AGI-TEST-001] Testing Claims of General Intelligence](../vol05_agi/AGI-TEST-001_testing_claims_of_general_intelligence.md): Testing coherence of world models.

## 11. References & Foundational Literature
1. Assran, M., Bardes, A., Fan, D., Garrido, Q., Howes, R., et al. (2025). *V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning*. arXiv preprint. arXiv:2506.09985 (30 authors). [Video joint-embedding model and zero-shot robot planning, as reported by the authors.]
2. Bruce, J., Dennis, M., Edwards, A., Parker-Holder, J., Shi, Y., Hughes, E., et al. (2024). *Genie: Generative Interactive Environments*. arXiv preprint. arXiv:2402.15391 (25 authors). [Approximately 11-billion-parameter foundation world model trained on unlabelled Internet video.]
3. Gurnee, W., & Tegmark, M. (2023). *Language Models Represent Space and Time*. arXiv preprint. arXiv:2310.02207. [Linear representations of space and time in Llama-2 models.]
4. Ha, D., & Schmidhuber, J. (2018). *World Models*. arXiv preprint. arXiv:1803.10122. [Generative world model with a controller trained in the model's dynamics.]
5. Hafner, D., Pasukonis, J., Ba, J., & Lillicrap, T. (2023). *Mastering Diverse Domains through World Models*. arXiv preprint. arXiv:2301.04104. [DreamerV3 and the Minecraft diamond claim, as stated by the authors.]
6. LeCun, Y. (2022). *A Path Towards Autonomous Machine Intelligence* (version 0.9.2, 2022-06-27). OpenReview position paper. https://openreview.net/forum?id=BZ5a1r-kVsf. [World-model-centred agent architecture. [VERIFICATION NEEDED: OpenReview page not directly retrieved; version and date confirmed only via secondary listings]]
7. Li, K., Hopkins, A. K., Bau, D., Viégas, F., Pfister, H., & Wattenberg, M. (2023). *Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task*. arXiv preprint. arXiv:2210.13382. [Board-state representation in an Othello-trained GPT-style model.]
8. Schrittwieser, J., Antonoglou, I., Hubert, T., Simonyan, K., Sifre, L., Schmitt, S., Guez, A., Lockhart, E., Hassabis, D., Graepel, T., Lillicrap, T., & Silver, D. (2020). *Mastering Atari, Go, chess and shogi by planning with a learned model*. Nature, 588(7839), 604–609. https://doi.org/10.1038/s41586-020-03051-4. [MuZero: planning with a learned model.]
9. Vafa, K., Chen, J. Y., Rambachan, A., Kleinberg, J., & Mullainathan, S. (2024). *Evaluating the World Model Implicit in a Generative Model*. arXiv preprint. arXiv:2406.03689. [Myhill-Nerode-based metrics finding less coherent implicit models than standard tests suggest.]

## 12. Evidential Classification Matrix
| Section / Claim | Evidence Category | Supporting Source | Confidence / Status |
| :--- | :--- | :--- | :--- |
| Learned models can support planning in games and simulation | Well Supported | Schrittwieser et al. (2020); Hafner et al. (2023) | High |
| DreamerV3 first to collect Minecraft diamonds from scratch | Preliminary (authors' claim) | Hafner et al. (2023) | Moderate |
| Generative interactive environment models exist as research systems | Preliminary | Bruce et al. (2024) | Moderate |
| V-JEPA 2 enables zero-shot robot planning in two labs | Preliminary (developer report) | Assran et al. (2025) | Moderate |
| LLMs contain implicit world models | Contested | Li et al. (2023); Gurnee & Tegmark (2023); Vafa et al. (2024) | Active debate |
| World models are necessary for AGI | Contested / Hypothetical | LeCun (2022) | Active debate |

## 13. Revision History
- **v0.1.0** (2026-10-15): Initial draft. Citations checked against arXiv/Crossref/publisher pages where marked; not independently fact-checked.