hi @lilyjge I'm looking at:
https://github.com/castorini/anserini/blob/master/docs/experiments-beir-fusion.md

Shouldn't "normalize" be a preprocessing step? In that you can normalize and then weight, but also weight unnormalized scores? And average is just weights with uniform weights, right?

So... shouldn't the organization be something like:

1. normalize score or not?
2. method?
   2.1 rrf
   2.2 weights (e.g., if not weights specified, assume equal weights = average)?

since rrf doesn't depend on weights, it should just be a no-op?

cc @FarmersWrap