Sunday, November 21, 2004

Letting Go

You just submitted your paper, working hard right up to the final deadline hour. Congratulations! Now what?

Forget about it. Oh you should distribute the paper: make it a technical report; email it to friends and family; put it on your web page and send it to an archive site. But don't keep working on it. All the tension you had getting the paper finished for the deadline needs a release. And no matter how much effort and time you continue to put into it, your paper will never be perfect.

Instead catch up on the todo's you've been ignoring. Say hi to the friends you have deserted while you bunkered down working. Take some time for you: catch a movie; eat a slow meal at a nice restaurant; take a walk.

Afterwards get back to research so you have another paper to write for the next deadline. You know you have successfully put that first paper out of your mind when you get caught off guard from the email from the PC chair letting you know your paper is

  • accepted: Great. Now you have to fix the paper up again for the proceedings deadline.
  • rejected: No problem. Now you have to fix the paper up again for the next conference deadline.
And it starts anew.


  1. do you really advocate arXiving or otherwise publicly releasing a submission ? do you think that it provides an adequate time stamp ? just curious

  2. I don't know what exactly you need for an "adequate time stamp," but the preprint archive at certainly has a number of submissions following the recent EUROCRYPT deadline. There is an issue here about anonymous submissions, but it's become International Association for Cryptologic Research policy that eprint postings should not predjudice reviews.

    -David Molnar

  3. this is a good point, and somewhat tricky. So I submit a paper to a conference implementing a double blind process. Is it ethical to then turn around and submit a preprint to arXiv or elsewhere, possibly negating the double blind process ?

  4. Re. Suresh's question --- of course it isn't ethical to do that, nor is it ethical to make subtle attempts to reveal one's identity when making a submission to a conf. with double-blind reviews.

    However, I think the whole idea of double-blind review is another instance of a conference (or a subset of the scholarly community) taking itself too seriously.

    Another example of a conference taking itself too seriously --- submissions in "proceedings format" --- this has gotten to an absurd point where for several conferences, the submissions are 10 or 12 pages in two-column, 9pt font with strange lines like "Copyright held by XYZ", etc. on the first page. Thankfully, sanity still prevails in pretty much all theory conferences (FOCS, STOC, CCC, TCC, etc.) --- submissions are usually required to respect a 10-page limit in 11pt or larger font, with reasonable margins all around...


  5. Strange, I never thought proceeding format was that much of a problem, considering they always provide the LaTeX files to do it correctly. (And that the result is much better looking than the LaTeX default styles.)

  6. Does the same advice hold true for grant proposals (sans distributing it?)

  7. Whether proceedings format looks better or not is a matter of personal taste, I guess --- I personally don't like anything below 11pt, personally prefer the CMR font to Times Roman (yes, I aware of "screen-readability" issues with bitmapped fonts, etc.)...
    There is a more objective reason why I dislike conference submissions in proceedings format: I don't feel it is worth spending cycles on properly formatting display math, eqnarrays, and, in the case of non-theory conferences, tables and figures, when I'd rather spend that time reading the paper one more time to check for awkward sentences and other miscellaneous typos.


  8. I'm very surprised by people saying that they like submissions to NOT be in proceeding format. I find this such a silly, ludicrous idea that I've toyed with the idea of simply boycotting conferences that won't allow me to submit something in (roughly) the final proceeding format.

    Why? Because I hate writing multiple versions of the same paper for no reason. I have to hack through my paper, do random things ranging from playing with margins, cutting out paragraphs, and moving random theorems into an appendix (based more on their size than their importance to the paper), just to have a submission version that turns out to be 10 "submission pages". Then I have to undo all these silly modifications at a later date to turn it into 10 "conference pages". Why couldn't I just send you what I thought the conference paper should look like in the first place? With the possible explanation of eyestrain for the committee (easily dealt with by various means), I just don't get why we have this process.

    Besides disliking this process as an author, I've also hated it as a PC member. I've seen papers that are 10 page submissions with 30+ page appendices. Again, this supposed "10-page limit" was helpful to me how?

  9. Siva, re: double blind submissions, this goes to a deeper point. you'd be surprised how bitterly accusation of politicizations are made in areas other than theory: double blind reviews appear to be one way of creating "a perception of fairness". Since justice must be "seen to have been done", perceptions are important.

    Theory tends to be relatively freer of this because to some degree, theory papers can be evaluated objectively, with the usual linear scale of faster/smaller/better. in other areas, such objective measurements are not always easy to make.

  10. Aside from formatting issues, I've seen conferences recently where the number of pages alotted for the proceedings version is less than the number allowed for submission, even if the formatting is the same for submission and final. You can then buy your way back up to the submitted number of pages, for $100-$150 per page, if your paper is accepted. If you, like me, feel that paying conferences leaves a bad taste in your mouth, then you have to spend time cutting the paper to fit.

    -David Molnar

  11. Crypto community, for one, has almost universally adopted anonymous submission process to conferences. (There are some weird exceptions like Financial Cryptography and TCC.) It is, I think, mostly useful for less-known authors since the most well-known people tend to advertise their papers anyway; I've seen Shamir talking about his paper in 3-4 different rump sessions before the paper itself was accepted. The less known author may also publish her paper on eprint, or opt to stay anonymous if she wants.

    I think it is better than in some other communities where submissions are not anonymous and reviewers only read the names of the authors and the abstract ("Oh, this paper comes from XYZ Research Labs. Abstract is interesting, let us accept it"), and I know that this is done actively in at least one community.

    - Helger L.

  12. I agree with Helger about the importance of anonymous submissions. Having been on program committees, and speaking only for myself, I am more prone to "give a paper some slack" when it is written by a well-known author and more likely to carefully check a paper for errors when it is written by an unknown author.

    I have heard two arguments to the above:
    (1) That I am wrong for doing so, and if I would correct my behavior then anonymous submissions would not be needed.

    This might be true (and I do try to catch myself when I do the above) but I think you cannot ignore the psychological bias which will always be present, even if only to a small degree. Also, who's to say that everyone else on the P.C. will act the same way?

    (2) The bias shown to known authors is a *good* idea, because clearly someone well-known is more likely to have a correct proof (say, of a long-standing open problem) than an unknown researcher.

    While I see the argument, I think it is unjustified and dangerous. Things also tend to not just work in this way: accepting papers is not just about finding papers whose proofs are correct, but about finding papers on interesting topics. And sometimes the "better-known" people seem (to some extent) to be the ones who decide what is interesting. (And this is why Suresh's comment about objectivity in theory conferences is not entirely true...)

  13. Firstly, I think submitting papers to archives is always a good idea, and no one should be discouraged from doing so. As mentioned above, this is indeed the convention even for CRYPTO conferences, where they have anonymous submissions.

    Regarding anonymous's comments above: you should make
    a clear separation between the importance of the results and the correctness of the results. In determining correctness, because of time constraints, I think it is reasonable to allow the reputation of the author to factor in to some extent, since you don't have time to verify everything. However, if I'm reviewing a paper and there's a point in the proof where I'm suspicious then, regardless of whether or not I know the author and whether or not they are famous, I'll try to get more information, either by looking at the appendix, getting a fuller version from the web, or emailing the author.

    Another point where I think it is reasonable to use the author's name is regarding related works. If I see a paper from an author who is a well known expert on some subject then I will tend to trust that what he wrote regarding previous works in this subject is accurate.

    When judging importance of the result and its potential contribution to the conference's audience, I try to use my own compass and knowledge, and there I think the author's name should not factor in. I hope that I am successful.

    In the few examples I recall of great papers that were rejected from conference (e.g., [GMR] paper on zero-knowledge and interactive proofs) I think the problem was not that the authors were not famous but that the papers were too novel to be appreciated.