publications - Substantial update of an arXiv paper

I put on arXiv several months ago a paper that has been submitted at the same time to a conference. The conference paper was rejected, and using the reviews, we have started improving the paper, and one thing leading to another, we have substantially changed the paper, to the point that the title has been changed. However, the new paper is similar in many points to the previous one, in particular about the technical results (the most important changes are rather on the presentation of the results).

We plan to submit in a near future the new version to a journal, and to also put this version on arXiv. My question is therefore the following one:

• Should we withdraw the previous version, and submit a new one?



• Should we replace the previous version by the new one, even though the title has changed? The arXiv paper is indexed by DBLP, would the DBLP citation be also updated?


• Should we keep the previous version, and the submit a new one?

Answer

You cannot completely withdraw a paper from arXiv; instead, the system can add a retraction notice on it. So, if you follow option #1, you'll end up with 1. an old paper, 2. with a retraction notice, 3. a new paper. It's a bit of a mess…

I would suggest the following:

1. 
   

   Consider your much-improved version as a new article: get it published, upload it on arXiv. In the arXiv version, you may refer to the older arXiv paper (“This article is based on the same results as …, but the presentation of results and the whole discuss have been reworked substantially”). Put it either in the introduction or, if you don't want to disrupt the flow of text, as a footnote or “banner” before the introduction.

   
   



2. 
   

   Most importantly: make a update to your old paper, for a version that links to the new paper. That way, you make it completely clear that a improved discussion is available.

   
   

This addressed the only issues I see, namely making sure that people can find your work, and that you're not suspected of duplicate publication (which, on an open platform like arXiv, doesn't make much sense anyway…).