Surprisingly enough, Mendeley correctly reads almost all the metadata from .pdf's. However, it doesn't work correctly with arXiv preprints: what I had is the arXiv link read as an article title (e.g. arXiv:0809.3331), and the blank field of publication name (where the journal title should be). Hope that the authors can fix it soon.
Another point is that at the moment there are only very few users of such services as Mendeley and 2collab are. However, I believe that in the nearest future using "academic social networks" will be as usual as, for instance, having a personal webpage at institute website nowadays.