Datasets:
ArXiv:
License:
Dataset Integrity
#3
by
outerform
- opened
Hi,
I noticed that in your paper, you mentioned there are 10 million Go game records in Video-GoBench. However, I found fewer than 600,000 .tar files in the dataset. I assume each .tar file corresponds to a single match, as there are no accompanying documents to clarify this. Could you please provide the missing files or clarify how to properly use the data?
Additionally, could you kindly specify the train/test split used in the experiments described in the paper?
Thank you!