# What a World Model Represents Is Three Questions

arXiv: https://arxiv.org/abs/2607.06640 · HTML: https://mlsanity.com/paper/2607.06640/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in World models for language agents, one of 11 directions in World models.
- Builds on 42 of its 59 references that are in the map.
- First posted on arXiv 7 Jul 2026.

## Placed in

- World models / [World models for language agents](https://mlsanity.com/wm/world-models-for-language-agents/)

## Builds on (most cited first, 30 of 42)

- [Shortcut Learning in Deep Neural Networks](https://mlsanity.com/paper/2004.07780/) (2020)
- [Dream to Control: Learning Behaviors by Latent Imagination](https://mlsanity.com/paper/1912.01603/) (2019)
- [Mastering Diverse Domains through World Models](https://mlsanity.com/paper/2301.04104/) (2023)
- [Mastering Atari with Discrete World Models](https://mlsanity.com/paper/2010.02193/) (2020)
- [Exact solutions to the nonlinear dynamics of learning in deep linear neural networks](https://mlsanity.com/paper/1312.6120/) (2013)
- [RMA: Rapid Motor Adaptation for Legged Robots](https://mlsanity.com/paper/2107.04034/) (2021)
- [Learning Invariant Representations for Reinforcement Learning without Reconstruction](https://mlsanity.com/paper/2006.10742/) (2020)
- [Reinforcement Learning with Unsupervised Auxiliary Tasks](https://mlsanity.com/paper/1611.05397/) (2016)
- [The Pitfalls of Simplicity Bias in Neural Networks](https://mlsanity.com/paper/2006.07710/) (2020)
- [Successor Features for Transfer in Reinforcement Learning](https://mlsanity.com/paper/1606.05312/) (2016)
- [Modern Koopman Theory for Dynamical Systems](https://mlsanity.com/paper/2102.12086/) (2021)
- [Gradient Starvation: A Learning Proclivity in Neural Networks](https://mlsanity.com/paper/2011.09468/) (2020)
- [FLAMBE: Structural Complexity and Representation Learning of Low Rank MDPs](https://mlsanity.com/paper/2006.10814/) (2020)
- [Emergence of Invariance and Disentanglement in Deep Representations](https://mlsanity.com/paper/1706.01350/) (2017)
- [Implicit Under-Parameterization Inhibits Data-Efficient Deep Reinforcement Learning](https://mlsanity.com/paper/2010.14498/) (2020)
- [What shapes feature representations? Exploring datasets, architectures, and training](https://mlsanity.com/paper/2006.12433/) (2020)
- [Objective Mismatch in Model-based Reinforcement Learning](https://mlsanity.com/paper/2002.04523/) (2020)
- [The Value Equivalence Principle for Model-Based Reinforcement Learning](https://mlsanity.com/paper/2011.03506/) (2020)
- [Hidden Parameter Markov Decision Processes: A Semiparametric Regression Approach for Discovering Latent Task Parametrizations](https://mlsanity.com/paper/1308.3513/) (2013)
- [Approximate information state for approximate planning and reinforcement learning in partially observed systems](https://mlsanity.com/paper/2010.08843/) (2020)
- [Denoised MDPs: Learning World Models Better Than the World Itself](https://mlsanity.com/paper/2206.15477/) (2022)
- [DR3: Value-Based Deep Reinforcement Learning Requires Explicit Regularization](https://mlsanity.com/paper/2112.04716/) (2021)
- [On the Foundations of Shortcut Learning](https://mlsanity.com/paper/2310.16228/) (2023)
- [Learning Task Informed Abstractions](https://mlsanity.com/paper/2106.15612/) (2021)
- [On The Effect of Auxiliary Tasks on Representation Dynamics](https://mlsanity.com/paper/2102.13089/) (2021)
- [Bridging State and History Representations: Understanding Self-Predictive RL](https://mlsanity.com/paper/2401.08898/) (2024)
- [Guaranteed Discovery of Control-Endogenous Latent States with Multi-Step Inverse Models](https://mlsanity.com/paper/2207.08229/) (2022)
- [Proper Value Equivalence](https://mlsanity.com/paper/2106.10316/) (2021)
- [Value Gradient weighted Model-Based Reinforcement Learning](https://mlsanity.com/paper/2204.01464/) (2022)
- [Action-Sufficient State Representation Learning for Control with Structural Constraints](https://mlsanity.com/paper/2110.05721/) (2021)

## Abstract (arXiv)

World models learn task-relevant information through many routes: observation reconstruction, recurrent state, temporal filtering, and explicit task supervision. Different routes can make different variables available. The same variable can also be available through several routes at once. When it is, looking at which route would increase the training loss most if removed does not tell you which route the model actually uses. The questions are reachability, whether a training signal can identify a task-relevant direction; admission, whether that direction is recoverable from the latent; and assignment, which eligible route carries it. We test them in environments with a known set of required coordinates. A direction cannot enter the latent unless some training signal can identify it. Reconstruction, recurrence, or filtering may already recover some of those coordinates; a reward or value head then has no residual direction to admit. For what remains, how many independent predictions the target supplies is how many coordinates install: one through four independent predictions admit one through four directions, including through the value head. Reachability is not admission: a temporal second-moment coefficient can remain absent under next-token prediction when accumulating it is a fraction of a percent of that loss, and a head that predicts the coefficient restores it. Assignment is a different test. Two routes that each carry the same variable when trained alone do not swap the carrier when we reverse which is more costly to remove. A recurrent model trained on a transformer's recorded sequences shows the same pattern. Near the point where the competing route is beginning to clear the probe threshold, independent training runs disagree. What a world model represents is therefore three questions: what information is reachable, what supervision admits, and which competing route carries it.