About Dataset Construction

[liuyijiang1994]

Thank you for your work. It has been very solid and made significant contributions in various aspects such as the dataset and experiments.

I have a few small questions that I would like to ask: In 4.1, it is mentioned that "On selected samples of the binary comparison task, we achieve an author-annotator agreement of 81.0% and annotator-annotator agreement of 70.0%."
How should we understand the "author-annotator agreement" and "annotator-annotator agreement"?
Both summaries are generated by the model, and in this case, humans only rank the two summaries. I can understand "annotator-annotator agreement" as the agreement between two annotators on the same sample.
But how should we understand "author-annotator agreement"?

I am looking forward to your reply. Thank you!