{
  "id": "2408.14837",
  "title": "Diffusion Models Are Real-Time Game Engines",
  "authors": "Dani Valevski, Yaniv Leviathan, Moab Arar, Shlomi Fruchter",
  "first_public": "2024-08-27",
  "first_public_source": "https://arxiv.org/abs/2408.14837",
  "venue": "ICLR 2025",
  "arxiv": "2408.14837",
  "paper_url": "https://arxiv.org/abs/2408.14837",
  "lens": [
    "interactive-worlds"
  ],
  "related_terms": [
    "GameNGen",
    "neural game engine",
    "real-time simulation",
    "diffusion renderer"
  ],
  "summary": "Presents GameNGen, a diffusion model that interactively simulates DOOM from previous frames and player actions.",
  "contribution": "Uses agent-collected gameplay, conditioning augmentation, and decoder fine-tuning to sustain autoregressive visual interaction.",
  "limitations": "The simulator is trained and evaluated on DOOM.",
  "model_output": "The next rendered game frame.",
  "action_interface": "DOOM player actions.",
  "state_memory": "Conditioning on a sequence of recent frames and actions; continued play feeds generated frames back into the model.",
  "relationship": "Interactive-simulation background cited by NeuralOS; NeuralOS also refers to its perceptual-evaluation setup.",
  "relationship_source": "https://arxiv.org/html/2507.08800v2",
  "author_names": [
    "Dani Valevski",
    "Yaniv Leviathan",
    "Moab Arar",
    "Shlomi Fruchter"
  ],
  "record_url": "https://neural-os.com/reading-list/papers/2408.14837/",
  "legacy_record_url": "https://neural-os.com/reading-list/papers/#paper-2408-14837",
  "citation_date": "2024-08-27",
  "json_url": "https://neural-os.com/reading-list/papers/2408.14837/record.json",
  "markdown_url": "https://neural-os.com/reading-list/papers/2408.14837/index.md",
  "bibtex_url": "https://neural-os.com/reading-list/papers/2408.14837/citation.bib",
  "csl_url": "https://neural-os.com/reading-list/papers/2408.14837/citation.csl.json"
}