One question about the paper: what is the "Direct-tune"?

[ffdffd]

Excellent work. In looking at your paper, it mentions "Direct-tune" which is described as:
"Direct-tune" refers to the approach of directly fitting the original video without outpainting training.
I don't understand it very well here, can I ask for a more specific explanation?

[JHLew]

Same question here, I failed to understand that part.
In addition, I am having a little struggle in reading, and wish for a detailed description of your wonderful work for better understanding, such as...

• which set of YoutubeVOS did you use in the experiments? All? or just the test set?
• I wonder what notations \bar{p}, \alpha_{lora}, exactly means...
• etc...
  Expecting to hear from you!