Learning to Theorize the World from Observation
cs.LG, cs.AI
Submitted: 2026-05-05
Updated: 2026-09-17
License: http://creativecommons.org/licenses/by/4.0/
The gist: What does it mean to understand the world? Contemporary world models often operationalize understanding as accurate future prediction in latent or observation space.
Terminology
Abstract
What does it mean to understand the world? Contemporary world models often operationalize understanding as accurate future prediction in latent or observation space. Developmental cognitive science, however, suggests a different view: human understanding emerges through the construction of internal theories of how the world works, even before mature language is acquired. Inspired by this theory-building view of cognition, we introduce Learning-to-Theorize, a learning paradigm for inferring explicit explanatory theories of the world from raw, non-textual observations. We instantiate this paradigm with the Neural Theorizer (NEO), a World Theory Model, that induces latent programs as a learned Language of Thought and executes them through a shared transition model. In NEO, a theory is represented as an executable, compositional program whose learned primitives can be systematically recombined to explain novel phenomena. Experiments show that this formulation enables explanation-driven generalization, allowing observations to be understood in terms of the programs that generate them.
Sources
- ARC Prize 2024: Technical Report
- ARC Prize 2025: Technical Report
- A Compressive-Expressive Communication Framework for Compositional Representations
- DreamCoder: Growing generalizable, interpretable knowledge with wake-sleep Bayesian program learning
- Genie: Generative Interactive Environments
- Differentiable Functional Program Interpreters
- On the Measure of Intelligence
- AdaWorld: Learning Adaptable World Models with Latent Actions
- Neural Turing Machines
- Hierarchical Programmatic Reinforcement Learning via Learning to Compose Programs
- Learning Latent Dynamics for Planning from Pixels
- Dream to Control: Learning Behaviors by Latent Imagination
- Mastering Atari with Discrete World Models
- LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels
- Mastering Diverse Domains through World Models
- Imagine the Unseen World: A Benchmark for Systematic Generalization in Visual World Models
- Auto-Encoding Variational Bayes
- Generalization without systematicity: On the compositional skills of sequence-to-sequence recurrent networks
- Learning to Synthesize Programs as Interpretable and Generalizable Policies
- Latent Action Pretraining from Videos
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: Local-Consistency Optimisation for Bayesian Neural Networks
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks