How to Carry a Good Benchmark?

There's too many papers of refrasing x as y, too many of them. If the information teretic interpretation is not givin me any new information, i don't want it. I'm tired of people willing that theory adds to a paper, yet most of the time, it either derais the message thatyoure trying to tell, or it's wrong, or it has nothing to do with the story. Information theory for machine learning is mostly useless. And here's a controversial opinion on this topic in its current form. It does not make sense in most cases,. Entropy does not makesense in most cases; given you have a gausian distribution and