{
  "title": "Genie action codes from video",
  "caption": "Genie trains action-conditioned dynamics with video tokens and latent actions inferred from unlabeled transitions. These latent actions are not calibrated robot commands. Author-redrawn mechanism from Bruce et al., without game footage and generated results. Source: Bruce et al. Playback time is illustrative and is not measured runtime.",
  "alt": "Genie flow: video tokens and inferred latent actions train dynamics, and interaction uses action codes without calibrated robot commands.",
  "sources": [
    "https://arxiv.org/abs/2402.15391"
  ],
  "limitations": [
    "This playback is an example. Playback duration is not measured runtime.",
    "Discrete latent-action IDs are not calibrated robot motor commands.",
    "The interaction trace uses a selected learned code. It does not infer control from a future frame.",
    "This diagram does not include source game footage or generated-result frames."
  ]
}