I'm now a PhD candidate mathematician, working towards understanding what really constitutes a paper as well. If you are considering going after a math PhD, I recommend finding an area you're interested in, finding papers in that field, and backtracking enough until you understand what's going on. I think math is very-well suited to this style of behavior.
For example, I'm interested in number theory, and I've recently been hearing a lot about automorphic forms and multiple Dirichlet Series. So I find this paper by Dr. Hoffstein. Some of it is understandable, some of it isn't. Going through it, I isolate 3 potential sources that might help me understand - the first references are frequently on background material, and here he references Selberg's work. Conveniently, my institution has access to electronic archives of his work, so that's great. It also references an Analytic Number Theory textbook by Iwaniec. Finally, I realize that this is largely built on this previous paper, with Dr. Hoffstein as one of the authors, from years before.
And then I can rinse, wash, and repeat. In this way, I both get an idea of what papers are like, how far removed they are from current material, and how advanced the math that goes into them are. A key aspect of this idea is that it's easier to go after particular research papers, theses, and dissertations than it is to find whole repositories that you can go through. So perhaps you should try a similar approach, suited to your interests.
As always, a good place to start is the arxiv.
On the other hand, it sounds like you're preparing to write a bachelors thesis, and that's of a different calibre. I suspect that your school has its own standards, and the best way to prepare for that is to simply do your best and ask your advisor lots of questions along the way. Ultimately, your school can't demand from an undergrad much more than it prepares you for (a vast difference from the life of overgrads, in my opinion).
As a last note, Harvey Mudd has a large archive of their undergrad math senior's theses here, and they might be what you're looking for if you don't like my previous idea.