The joke when I was in academia was that there were two ways to get lots of citations: publish a state-of-the-art model that gets cited because it's state-of-the-art, or publish a bad model that gets cited a lot because everyone wants to show how much they beat it by.