Comment by maxov

What a strange recommendation. I do research in CS theory and machine learning and I try to find arXiv preprints when I can, they are usually more complete than conference versions of papers. If you stick to papers from authors you know or to those from well-known conferences, arXiv is often times the best source.