{
  "title": "DreamDojo: pretrain, adapt and distill",
  "caption": "The author made this diagram from the source figure. Human transitions supply proxy actions. Target-robot adaptation grounds the command interface before autoregressive distillation. The training phases are different. The playback is an example. It does not show measured training time or generated-video performance.",
  "alt": "The training flow uses human-video latent actions for pretraining, robot data for action grounding, then autoregressive distillation.",
  "sources": [
    "https://arxiv.org/abs/2602.06949",
    "https://dreamdojo-world.github.io/"
  ],
  "limitations": [
    "The playback is an example. Playback time is not measured runtime.",
    "The diagram does not show robot photographs, generated frames or measured task performance."
  ]
}