Missing samples

[n0brainn0gain]

Hello, Agibot-world!

You write the following sentence in your paper on p. 12:

"To construct a comprehensive and diverse evaluation set, we uniformly sample 100 video instances per
task. Each task is decomposed into fine-grained primitive actions, ensuring one-to-one correspondence
between video segments, sub-action labels, and textual descriptions."

But in gt_dataset.tar one can find only 21 prompts. Where we can find all of the missing videos?

Thank you in advance.

[ryancll]

@n0brainn0gain Thank you for your interest! We’ve released a small verification subset of our ground-truth dataset (21 samples, 7 task categories) for fast testing. The full version will follow in the near future.

[n0brainn0gain]

@ryancll Thank you for your previous response regarding the verification subset of the ground-truth dataset. We truly appreciate your efforts in making this available for testing.

Could you kindly provide an update on when we might expect the full set of prompts to be released? It would be incredibly helpful for our evaluation and further experimentation.

Thanks in advance for your time and support! Looking forward to hearing from you.

[buchuoer]

Are there any plans to open - source it recently?

[B0B8K1ng]

any updates?