Ways to resolve Popularity Bias in warp or bpr?

I'm currently working on a system to rank different restaurants. Currently, I see the effect of popularity bias in restaurant ranking like most frequent ordered restaurants is mostly being recommended.

Is there any way to recommend items from the longtail item list?
Do we have any tunable mechanism for controlling the trade-off between accuracy and coverage?

No file, test, or entry point is named. Start by reviewing how the Python API exposes WARP and BPR ranking, then determine whether popularity or long-tail coverage controls already exist. Done should include a clearly defined accuracy-versus-coverage mechanism and tests or documentation showing how to use it.

Written by the indexing model from the issue text.