mlsanityData to 14 Oct 2026

Context and Diversity Matter: The Emergence of In-Context Learning in World Models

arXiv 2509.22353 ↗First posted 26 Sep 2025cs.LG cs.AIcited by 3

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

Abstract (arXiv)

The capability of predicting environmental dynamics underpins both biological neural systems and general embodied AI in adapting to their surroundings. Yet prevailing approaches rest on static world models that falter when confronted with novel or rare configurations. We investigate in-context learning (ICL) of world models, shifting attention from zero-shot performance to the growth and asymptotic limits of the world model. Our contributions are three-fold: (1) we formalize ICL of a world model and identify two core mechanisms: environment recognition (ER) and environment learning (EL); (2) we derive error upper-bounds for both mechanisms that expose how the mechanisms emerge; and (3) we empirically confirm that distinct ICL mechanisms exist in the world model, and we further investigate how data distribution and model architecture affect ICL in a manner consistent with theory. These findings demonstrate the potential of self-adapting world models and highlight the key factors behind the emergence of EL/ER, most notably the necessity of long context and diverse environments.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Placed in World models for language agents, one of 11 directions in World models. Builds on 64 of its 83 references that are in the map. First posted on arXiv 26 Sep 2025.

Builds on

64 of 83 references.

Closest ancestors · 1

Same direction, newest first

  1. 2020Language Models are Few-Shot Learners28 May

Other references · 63

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  2. 2019Dream to Control: Learning Behaviors by Latent Imaginationcited by 1,653
  3. 2018World Modelscited by 1,513
  4. 2020Linformer: Self-Attention with Linear Complexitycited by 1,474
  5. 2019Mastering Atari, Go, Chess and Shogi by Planning with a Learned Modelcited by 1,353
  6. 2018Learning Latent Dynamics for Planning from Pixelscited by 1,264
  7. 2021An Explanation of In-context Learning as Implicit Bayesian Inferencecited by 938
  8. 2017AI2-THOR: An Interactive 3D Environment for Visual AIcited by 869
  9. 2018Recurrent World Models Facilitate Policy Evolutioncited by 785
  10. 2022What Can Transformers Learn In-Context? A Case Study of Simple Function Classescited by 715
  11. 2023GAIA-1: A Generative World Model for Autonomous Drivingcited by 637
  12. 2023Gated Linear Attention Transformers with Hardware-Efficient Trainingcited by 514
38 more
  1. 2023TD-MPC2: Scalable, Robust World Models for Continuous Controlcited by 512
  2. 2016RL2: Fast Reinforcement Learning via Slow Reinforcement Learningcited by 503
  3. 2019Model-Based Reinforcement Learning for Ataricited by 502
  4. 2016Reinforcement Learning with Unsupervised Auxiliary Taskscited by 404
  5. 2022DayDreamer: World Models for Physical Robot Learningcited by 400
  6. 2022ProcTHOR: Large-Scale Embodied AI Using Procedural Generationcited by 380
  7. 2024Vista: A Generalizable Driving World Model with High Fidelity and Versatile Controllabilitycited by 347
  8. 2023Driving into the Future: Multiview Visual Forecasting and Planning with World Model for Autonomous Drivingcited by 319
  9. 2022Data Distributional Properties Drive Emergent In-Context Learning in Transformerscited by 309
  10. 2024Diffusion for World Modeling: Visual Details Matter in Ataricited by 300
  11. 2016Deep Visual Foresight for Planning Robot Motioncited by 285
  12. 2024Navigation World Modelscited by 272
  13. 2024RoboDreamer: Learning Compositional World Models for Robot Imaginationcited by 195
  14. 2022In-context Reinforcement Learning with Algorithm Distillationcited by 190
  15. 2025GAIA-2: A Controllable Multi-View Generative World Model for Autonomous Drivingcited by 185
  16. 2024Enhancing End-to-End Autonomous Driving with Latent World Modelcited by 166
  17. 2023What In-Context Learning "Learns" In-Context: Disentangling Task Recognition and Task Learningcited by 152
  18. 2022Model-Based Imitation Learning for Urban Drivingcited by 150
  19. 2023Pretraining task diversity and the emergence of non-Bayesian in-context learning for regressioncited by 146
  20. 2023Supervised Pretraining Can Learn In-Context Reinforcement Learningcited by 139
  21. 2021Discovering and Achieving Goals via World Modelscited by 131
  22. 2021Pathdreamer: A World Model for Indoor Navigationcited by 118
  23. 2023Copilot4D: Learning Unsupervised World Models for Autonomous Driving via Discrete Diffusioncited by 106
  24. 2022General-Purpose In-Context Learning by Meta-Learning Transformerscited by 105
  25. 2023STORM: Efficient Stochastic Transformer based World Models for Reinforcement Learningcited by 89
  26. 2024In-Context Language Learning: Architectures and Algorithmscited by 89
  27. 2023In-Context Learning through the Bayesian Prismcited by 83
  28. 2023The Transient Nature of Emergent In-Context Learning in Transformerscited by 76
  29. 2024Gated Slot Attention for Efficient Linear-Time Sequence Modelingcited by 76
  30. 2023Bird's-Eye-View Scene Graph for Vision-Language Navigationcited by 72
  31. 2024Learning and Leveraging World Models in Visual Representation Learningcited by 72
  32. 2023TrafficBots: Towards World Models for Autonomous Driving Simulation and Motion Predictioncited by 56
  33. 2024Competition Dynamics Shape Algorithmic Phases of In-Context Learningcited by 54
  34. 2023SPOC: Imitating Shortest Paths in Simulation Enables Effective Navigation and Manipulation in the Real Worldcited by 48
  35. 2024Dream to Manipulate: Compositional World Models Empowering Robot Imitation Learning with Imaginationcited by 45
  36. 2022Evaluating Long-Term Memory in 3D Mazescited by 43
  37. 2025WMNav: Integrating Vision-Language Models into World Models for Object Goal Navigationcited by 43
  38. 2024Mastering Memory Tasks with World Modelscited by 38

50 most cited of 63 All 63 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 1; 2026: 2.

3 papers.

  1. 2025Benchmarking World-Model Learning with Environment-Level Queriescited by 10
  2. 2026SyncWorld: Visual Calibration Enables World Models as Zero-Shot Simulatorscited by 1
  3. 2026Learning What to Recall: Adaptive Multi-Cue Episodic Memory for World Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Grounded world models in biological organisms and future embodied AI
  2. 2025SmallWorlds: Assessing Dynamics Understanding of World Models in Isolated Environments
  3. 2026Shaping Persistent Representations from Independent Interactions
  4. 2025World-in-World: World Models in a Closed-Loop World
  5. 2026Research on World Models Is Not Merely Injecting World Knowledge into Specific Tasks
  6. 2026Task Ecologies and the Evolution of World-Tracking Representations in Large Language Models
  7. 2026What Do World Models Learn in RL? Probing Latent Representations in Learned Environment Simulators
  8. 2025From Masks to Worlds: A Hitchhiker's Guide to World Models