> Their approach was basically to look at publications from the previous conference, apply some delta that they had an advantage over other teams and publish th

> Their approach was basically to look at publications from the previous conference, apply some delta that they had an advantage over other teams and publish this in the next round. I think that's it for most PhD students. Look at recent papers from your advisor (or their co-authors). Once you understand that specialized area well-enough, you should be able to find an epsilon that can be added, or to apply some results to a different use case. A 2-3 papers in somewhat decent conferences should be enough to be able to graduate. Most people won't solve interesting open problems. They will find questions that nobody has asked before, so they can be the first one to answer it.