Question about RoboCasa video world model finetuning

[ZihaoLu001]

Thanks for releasing this interesting project!

While reading the DREAMGEN paper, I noticed that the video world model is fine-tuned on RoboCasa human demonstrations (e.g., Section 2.1 and Appendix D). However, I could not find the corresponding RoboCasa finetuning scripts or checkpoints in the public GR00T-Dreams repository.

Could you please clarify whether the RoboCasa video world model finetuning was done using an internal pipeline, or if this part of the code is planned to be released?

Thanks!