Datasets:

---

i-need-sleep

/

musicprefs

like 3

Hello,

In the released dataset, is the first audio file (the one listed above) always the human-preferred (win) output, or is the ordering randomized?

(example)

And, if ordering is randomized, do you plan to make the human preference annotations publicly available in the future, or will they remain internal?

Thank you very much for your clarification and for your contribution in building this valuable dataset! :)