Disclaimer: this post will mostly be a rant. The reasons for it should become sufficiently clear as I go over my experience in the past few months.
For context, I have been working on a paper for the past year or so. Said paper involved running a few thousands experimental configurations over different baselines, then reporting said experiments in a paper that was initially submitted to VLDB2024 (end of January this year). Said paper was rejected at VLDB 2024, submitted to SIGMOD2025, and was rejected again.
Over the course of only a few months, I had the privilege of experiencing (very courteous) academic blackmailing, extremely negative reviews, burnout, overtime, and possibly the most bullshit rejection I have heard of.
Let’s start from top, with the academic blackmail. So, VLDB is a single blind conference, which means that the authors are not aware of who the reviewers are, while the reviewers can instead see the names and affiliations of the authors.
This is less annoying to manage than double blind (which instead involve anonymizing everything), but has the fun side effect of receiving emails from the reviewers, who might nicely ask to please add their recent papers to yours in a completely non-threatening way by reaching out in private to one of the authors. Bonus points for asking to add references to papers whose code is not available to run and test.
This leads me directly into the rejection. I have my fair share of rejected papers – though by this point, most of my rejections are from this singular paper –, and I am keenly aware of how much of an impact “rolling the right reviewers” has on the final outcome. This rejection was especially fun, however, because it came only after a revision period during which I had to re-run most of the experiments (hence the burnout). But as a man once said, “I’m not done yet”: the reasoning for the rejection was that “we did not test a specific baseline, and said baseline works on my machine” (paraphrased). The kicker? The code for the baseline was not available online, and the reviewer did not provide it to us so that we could test it. As a result, the area chair agreed with it, and the paper was rejected after the revision step.
What was even more of a kick in the nether regions was that we even managed to change the mind of the one reviewer that had marked his review as “weak reject, and I am not going to change my mind”. Unfortunately, we also managed to change the mind of one of the “weak accept” reviews to “reject”, and that somehow stuck.
Something interesting to note is that, soon after the paper was rejected, the arXiv version received a new citation. This citation was by one of the authors of the baseline. While I am not one to engage in conspirationsm, the coincidence was a bit too perfect for me to ignore.
While I only briefly mentioned the burnout, it was bad. I got sick after every deadline, and I think I developed some degree of PTSD from looking at terminal screens waiting for experimental runs to end. Not the best, especially considering it did not result in success (yet, at least). That’s not good for morale either.
Lessons learned
What is the summary of these experiences? 1) The current reviewing system is broken (which is something literally anyone in the field will confirm); 2) academia is rife with politics just like everything else; 3) effort is not likely to be rewarded, and your fate lies less in your hands and in your results, than in the ethereal roll of the dice that will assign a reviewer rather than another in any given round of paper submission.
The result? I will be looking for my next position in industry, rather than academia. Stay tuned for when I will write the equivalent post on my industry misgivings once I have enough experience in that.