What exactly does Microsoft's WHAM/Muse release offer for gameplay ideation, and how do the Nature paper and the model card differ?

WHAM is a single-game model (Bleeding Edge), 300x180 at 10 Hz, with a 1-second context window, built from a user study of game creatives and evaluated on consistency (FVD), diversity (Wasserstein) and persistency (blinded human labels). The released 200M and 1.6B checkpoints are research-licensed (non-commercial, no distribution), not real time, and domain-bound. The model card's 'Limited context length (10s)' conflicts with the paper's 1 s of context; the card's own '10 pairs' agrees with the paper. Curator qualifications: 'intended for academic research purposes only'; 'Context length: 10 (observation, action) pairs / 5560 tokens'; data at 10Hz; 'Limited context length (10s)'; 'licensed under the Microsoft Research License'. Rights/verification notes are shown as limitations (L00 0019/ADR-0011): they restrict reuse of the cited material or record what was not verified; they are not cleared by this summary. Limitations: [component_reuse_restriction] Microsoft Research License text (LICENSE.md) not read by curator; distribution/output-use terms per producer only. Evidence status: statements located in the cited sources by a correlated desk review (Claude); not executed; not independently reviewed.

{
  "citations": [
    {
      "url": "https://www.nature.com/articles/s41586-025-08600-3",
      "note": "HF microsoft/wham README at 96339932 lines 169, 183, 204, 225-227"
    },
    {
      "url": "https://huggingface.co/microsoft/wham/blob/963399322b737416569ab62e269b1d6d02b00e8f/README.md",
      "note": "HF microsoft/wham README at 96339932 lines 169, 183, 204, 225-227"
    }
  ]
}