What should I do if I find a mistake after I submitted the camera-ready paper?

Question

I recently got one paper accepted. Then I added one paragraph given the reviewers' comments and run more experiments for this section. The problem is that I don't have much time to run the experiments so I did not fully understand the phenomenon behind the experiments. And I used a set of parameters that makes the baselines perform fairly low. After submission, I realized one baseline could perform much better if I use another set of parameters. I re-run the experiments and found that my concern is true.

Now I plan to release the code. I believe that someone would find out this problem if they take some time to run the experiments and tune the parameters. The deadline for camera-ready has passed already. What could I do to fix this? I don't want to be criticized that I intentionally make the baseline perform low by selecting hyperparameters. Could I correct this through arxiv? For example, I upload an updated version of this work on arxiv? This section even has nothing to do with the paper's acceptance because it is added after the paper got accepted. I just want to make the paper more solid while making new mistakes. Really appreciate the opinions.

Answer 1

Contact the publisher immediately. Your odds of making changes to an accepted paper decrease drastically once the publisher starts typesetting it. Unless you are a famous author which the publisher is especially incentivized to please, it is extremely unlikely you'll be able to make changes on the scale you describe after you receive the first proofs.

Beyond that, you'll just have to work with the publisher over what to do next. You could for example ask them to send it back to the editor if you're concerned about the changes you made - but no matter what happens, the first step (contact the publisher) is time sensitive, and you must do it anyway.

Answer 2

First, you should take this as a lesson not to hastily add things to papers.

Second, as Allure says, you can contact the editor, and see if it is still possible to fix a mistake. It might or might not be. If it is, you're in luck. If not, here are a few other options:

• If you think the mistake's important enough to fix, publish an corrigendum later.

• You can always post a corrected version/corrigendum on the arXiv or your website, assuming you are in a field covered by the arXiv. Depending on the culture in your field, the arXiv version may or may not be more commonly viewed than the published version, so this may or may not be effective.

• When you give talks or have discussions about the relevant work, you can point out the error.