mlsanityData to 7 Oct 2026

Robust agents learn causal world models

arXiv 2402.10877 ↗First posted 16 Feb 2024cs.AI cs.LGcited by 81

On the map: World models → World models for language agents

also in 1 more
Abstract (arXiv)

It has long been hypothesised that causal reasoning plays a fundamental role in robust and general intelligence. However, it is not known if agents must learn causal models in order to generalise to new domains, or if other inductive biases are sufficient. We answer this question, showing that any agent capable of satisfying a regret bound under a large set of distributional shifts must have learned an approximate causal model of the data generating process, which converges to the true causal model for optimal agents. We discuss the implications of this result for several research areas including transfer learning and causal inference.

Key facts. Cited by 81 papers in the map (data through 7 Oct 2026). Placed in World models for language agents, one of 11 directions in World models. Builds on 46 of its 83 references that are in the map. First posted on arXiv 16 Feb 2024.

Builds on

46 of 83 references.

Closest ancestors · 1

Same direction, newest first

  1. 2020Language Models are Few-Shot Learners28 May

Other references · 45

  1. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,857
  2. 2019Invariant Risk Minimizationcited by 1,832
  3. 2018Relational inductive biases, deep learning, and graph networkscited by 1,341
  4. 2018On the Spectral Bias of Neural Networkscited by 1,321
  5. 2016Group Equivariant Convolutional Networkscited by 1,154
  6. 2019A Comprehensive Survey on Transfer Learningcited by 1,154
  7. 2017Counterfactual Explanations without Opening the Black Box: Automated Decisions and the GDPRcited by 1,086
  8. 2017Counterfactual Fairnesscited by 802
  9. 2022A Generalist Agentcited by 802
  10. 2021Generalizing to Unseen Domains: A Survey on Domain Generalizationcited by 762
  11. 2021Domain Generalization: A Surveycited by 760
  12. 2017Sim-to-Real Transfer of Robotic Control with Dynamics Randomizationcited by 616
33 more
  1. 2017Zero-Shot Learning -- A Comprehensive Evaluation of the Good, the Bad and the Uglycited by 536
  2. 2021Towards Out-Of-Distribution Generalization: A Surveycited by 431
  3. 2022Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Taskcited by 417
  4. 2018A Survey of Unsupervised Deep Domain Adaptationcited by 361
  5. 2023Causal Reasoning and Large Language Models: Opening a New Frontier for Causalitycited by 348
  6. 2014In Search of the Real Inductive Bias: On the Role of Implicit Regularization in Deep Learningcited by 339
  7. 2023Language Models Represent Space and Timecited by 306
  8. 2020A Brief Review of Domain Adaptationcited by 287
  9. 2020Inductive Biases for Deep Learning of Higher-Level Cognitioncited by 273
  10. 2021D'ya like DAGs? A Survey on Structure Learning and Causal Discoverycited by 209
  11. 2018Why do deep convolutional networks generalize so poorly to small image transformations?cited by 198
  12. 2019Causality matters in medical imagingcited by 193
  13. 2021Maximum Entropy RL (Provably) Solves Some Robust RL Problemscited by 169
  14. 2021A Survey of Zero-shot Generalisation in Deep Reinforcement Learningcited by 164
  15. 2023Causal Parrots: Large Language Models May Talk Causality But Are Not Causalcited by 146
  16. 2021Goal Misgeneralization in Deep Reinforcement Learningcited by 122
  17. 2021Acquisition of Chess Knowledge in AlphaZerocited by 120
  18. 2022Goal Misgeneralization: Why Correct Specifications Aren't Enough For Correct Goalscited by 120
  19. 2021Can Language Models Encode Perceptual Structure Without Grounding? A Case Study in Colorcited by 118
  20. 2018Theoretical Impediments to Machine Learning With Seven Sparks from the Causal Revolutioncited by 66
  21. 2021Agent Incentives: A Causal Perspectivecited by 51
  22. 2013Strong Completeness and Faithfulness in Bayesian Networkscited by 44
  23. 2023Honesty Is the Best Policy: Defining and Mitigating AI Deceptioncited by 40
  24. 2014Distinguishing cause from effect using observational data: methods and benchmarkscited by 33
  25. 2018Towards Formal Definitions of Blameworthiness, Intention, and Moral Responsibilitycited by 32
  26. 2017Causally Regularized Learning with Agnostic Data Selection Biascited by 29
  27. 2022Counterfactual harmcited by 29
  28. 2022Discovering Agentscited by 28
  29. 2024The Reasons that Agents Act: Intention and Instrumental Goalscited by 23
  30. 2018Causal Inference via Kernel Deviance Measurescited by 22
  31. 2022Selection, Ignorability and Challenges With Causal Fairnesscited by 16
  32. 2021A Topological Perspective on Causal Inferencecited by 10
  33. 2022Promises and Challenges of Causality for Ethical Machine Learningcited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 19; 2025: 33; 2026: 29.

81 papers.

Sort:

By area:

  1. 2024Mechanistic Interpretability for AI Safety -- A Reviewcited by 415
  2. 2024The Platonic Representation Hypothesiscited by 362
  3. 2025AI Agents vs. Agentic AI: A Conceptual Taxonomy, Applications and Challengescited by 167
  4. 2024Making Large Language Models into World Models with Precondition and Effect Knowledgecited by 21
  5. 2025Agentic AI Needs a Systems Theorycited by 19
  6. 2025A Survey on Large Language Models for Automated Planningcited by 17
  7. 2024Identifiable Exchangeable Mechanisms for Causal Structure and Representation Learningcited by 16
  8. 2025A Step Toward World Models: A Survey on Robotic Manipulationcited by 15
  9. 2024CLEAR: Can Language Models Really Understand Causal Graphs?cited by 14
  10. 2024Three Dogmas of Reinforcement Learningcited by 14
  11. 2025General agents contain world modelscited by 14
  12. 2025Towards Interpretable Deep Generative Models via Causal Representation Learningcited by 12
38 more
  1. 2024SPARTAN: A Sparse Transformer World Model Attending to What Matterscited by 10
  2. 2025How Causal Abstraction Underpins Computational Explanationcited by 10
  3. 2025When and How Does CLIP Enable Domain and Compositional Generalization?cited by 9
  4. 2024Language Agents Meet Causality -- Bridging LLMs and Causal World Modelscited by 8
  5. 2025When Do Neural Networks Learn World Models?cited by 8
  6. 2024Measuring Goal-Directednesscited by 7
  7. 2025A Mathematical Philosophy of Explanations in Mechanistic Interpretability -- The Strange Science Part I.icited by 6
  8. 2025Preference Learning for AI Alignment: a Causal Perspectivecited by 6
  9. 2025Collaborate, Deliberate, Evaluate: How LLM Alignment Affects Coordinated Multi-Agent Outcomescited by 6
  10. 2026Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world modelscited by 6
  11. 2024Causality from Bottom to Top: A Surveycited by 5
  12. 2025Enhancing LLM Agent Safety via Causal Influence Promptingcited by 5
  13. 2024Are Biological Systems More Intelligent Than Artificial Intelligence?cited by 4
  14. 2025Towards Empowerment Gain through Causal Structure Learning in Model-Based RLcited by 4
  15. 2025CausalMan: A physics-based simulator for large-scale causalitycited by 4
  16. 2025Linear Spatial World Models Emerge in Large Language Modelscited by 4
  17. 2025Curious Causality-Seeking Agents Learn Meta Causal Worldcited by 4
  18. 2025CausalARC: Abstract Reasoning with Causal World Modelscited by 4
  19. 2026Overcoming Statistical Bias in Action-Controllable World Modelscited by 4
  20. 2025Safety Must Precede the Deployment of Open-Ended AIcited by 3
  21. 2025The Limits of Predicting Agents from Behaviourcited by 3
  22. 2025A Principle of Targeted Intervention for Multi-Agent Reinforcement Learningcited by 3
  23. 2026YoCausal: How Far is Video Generation from World Model? A Causality Perspectivecited by 3
  24. 2025Causality Is Key to Understand and Balance Multiple Goals in Trustworthy ML and Foundation Modelscited by 2
  25. 2025The Relativity of Causal Knowledgecited by 2
  26. 2025Learning Natural Language Constraints for Safe Reinforcement Learning of Language Agentscited by 2
  27. 2025Learning Local Causal World Models with State Space Models and Attentioncited by 2
  28. 2026CofactVLA: Deconfounding Vision-Language-Action Models via Counterfactual Interventioncited by 2
  29. 2024Imitation Game: A Model-based and Imitation Learning Deep Reinforcement Learning Hybridcited by 1
  30. 2024Why Is Anything Conscious?cited by 1
  31. 2024Exploring the Learning Capabilities of Language Models using LEVERWORLDScited by 1
  32. 2024Factored space models: Towards causality between levels of abstractioncited by 1
  33. 2025Trust Me, I Know the Way: Predictive Uncertainty in the Presence of Shortcut Learningcited by 1
  34. 2025Towards Causal Model-Based Policy Optimizationcited by 1
  35. 2025Efficient and Generalizable Environmental Understanding for Visual Navigationcited by 1
  36. 2026Linear-LLM-SCM: Benchmarking LLMs for Coefficient Elicitation in Linear-Gaussian Causal Modelscited by 1
  37. 2026Calculating Mutual Information between a Reward Maximizer and its Environmentcited by 1
  38. 2026What Capable Agents Must Know: Selection Theorems for Robust Decision-Making under Uncertaintycited by 1

50 most cited of 81 All 81 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026An Introduction to Causal Reinforcement Learning
  2. 2026Causality as the Statistical Conscience of Artificial Intelligence: From Pearl's Ladder to Trustworthy Machines
  3. 2024Why Online Reinforcement Learning is Causal
  4. 2025Causal Reflection with Language Models
  5. 2022Generalizing Goal-Conditioned Reinforcement Learning with Variational Causal Reasoning
  6. 2025Towards Error-Centric Intelligence II: Energy-Structured Causal Models
  7. 2024Boosting Efficiency in Task-Agnostic Exploration through Causal Knowledge
  8. 2026Why LLMs Fail at Causal Discovery and How Interventional Agents Escape