{
  "title": "Search quantities in MuZero",
  "caption": "MuZero uses h for history encoding, g for latent transitions and rewards, and f for policy/value predictions in search. After environment actions, h encodes new observations. Author-redrawn mechanism from Schrittwieser et al. The source figure keeps the original board snapshots, full search tree and training-unroll panel. Source: Schrittwieser et al. Playback time is illustrative and is not measured runtime.",
  "alt": "MuZero search: history encoding supplies latent dynamics, reward, policy and value predictions before action and new observations.",
  "sources": [
    "https://arxiv.org/abs/1911.08265"
  ],
  "limitations": [
    "This playback is an example. Playback duration is not measured runtime.",
    "The diagram does not include search-visit statistics, full selection/backup rules and training unroll.",
    "This expansion does not show a fixed search depth or measured runtime."
  ]
}