I’ve met a few very successful PhD students, or undergrads/MS students who were very successful at publishing. In general, they tend to have some combination of the following qualities
1. excitement about the problems. This seems to be hugely helpful.
2. focus. For the most successful student in our lab, I notice that while he chats with us, the percentage of the time he spends glued to the computer debugging is higher than anyone else.
3. advisors pushing hard can make students look brilliant
4. often they have strong technical skills in at least one area. In computer science/AI, it could be HPC, systems programming, some area of math, or domain knowledge of an area where one can apply CS or ML.
5. Strong writing skills and the ability to ‘sell’ in their writing, making claims that are strong but at the same time not unreasonable.
Note that of these qualities for success, natural ‘brilliance’ only plays a role in developing the fourth one more quickly. That said, one can’t discount the fourth one as it can be hugely helpful.