{
  "id": "1803.10122",
  "title": "World Models",
  "authors": "David Ha, Jürgen Schmidhuber",
  "first_public": "2018-03-27",
  "first_public_source": "https://arxiv.org/abs/1803.10122",
  "venue": "arXiv",
  "arxiv": "1803.10122",
  "paper_url": "https://arxiv.org/abs/1803.10122",
  "lens": [
    "interactive-worlds"
  ],
  "related_terms": [
    "world model",
    "recurrent dynamics",
    "latent imagination",
    "model-based reinforcement learning"
  ],
  "summary": "Combines learned visual compression and recurrent dynamics with a controller, including training an agent inside a generated environment.",
  "contribution": "Demonstrates how a compact learned environment representation can support policy learning and transfer from a simulated dream to the original task.",
  "limitations": "The model compresses observations into a latent representation for control.",
  "model_output": "Predicted latent observations, with an image decoder and task-dependent termination predictions.",
  "action_interface": "Actions from the controller in the studied reinforcement-learning environments.",
  "state_memory": "A recurrent model summarizes the history of latent observations and actions.",
  "relationship": "Cited by NeuralOS as background for learned simulation and agent training.",
  "relationship_source": "https://arxiv.org/html/2507.08800v2",
  "author_names": [
    "David Ha",
    "Jürgen Schmidhuber"
  ],
  "record_url": "https://neural-os.com/reading-list/papers/1803.10122/",
  "legacy_record_url": "https://neural-os.com/reading-list/papers/#paper-1803-10122",
  "citation_date": "2018-03-27",
  "json_url": "https://neural-os.com/reading-list/papers/1803.10122/record.json",
  "markdown_url": "https://neural-os.com/reading-list/papers/1803.10122/index.md",
  "bibtex_url": "https://neural-os.com/reading-list/papers/1803.10122/citation.bib",
  "csl_url": "https://neural-os.com/reading-list/papers/1803.10122/citation.csl.json"
}