I believe you should never include "Submitted to Journal of Blah"; it conveys no information and it's tacky. (And just because Terry Tao [does it][1] doesn't make it okay.) I won't post a new version of a paper just because it has been accepted. However, if I do upload a newer version, I'll include in the comments "Final version, to appear in Journal of Blah" (here is [an example][2] from one of my papers). Once the paper is published, with page numbers and everything (which can often be years -- my paper above was accepted in April 2010, published January 2013), there is a [form](http://arxiv.org/help/jref) you can use to add the journal reference and DOI to the arXiv page. Conveniently, this does *not* generate a new version, so there's no reason not to (you can see this on my paper above -- the reference and DOI were added last week, but no new version of the paper was generated). [1]: http://arxiv.org/abs/1211.6372 [2]: http://arxiv.org/abs/0905.2360