There is a quiet panic in the question posted by a researcher who wants to use numbered citations at ICLR, despite explicit instructions for Author Year format. The question asks, point blank, whether deviating from the style guide will trigger a straight desk rejection. It is a small question, but it carries a larger weight. We have seen how easily small formatting choices can ripple into bigger outcomes, especially when the underlying goal is to keep a model or a dataset clean and reliable. The same instinct that makes a researcher ask about citation format is the instinct that drives careful data hygiene, whether you are cleaning data before it skews your model or deciding how to handle edge cases in real-world computer vision deployments. Rules are not arbitrary, but they are also not the point. The point is whether the work is understood.
Let us be direct: submitting against the stated instructions is a gamble, and it is not a smart one. The instructions exist because reviewers read a lot of papers, and consistency helps them focus on the science rather than the formatting. A numbered citation list is not a moral failing, but it signals something. It signals that you either did not read the requirements or you chose to ignore them. Neither is a strong opening move in a venue where first impressions matter. The researcher who asked the question likely knows this. They are looking for a loophole, a sign that the rule is bendable, or maybe just a story of someone who got away with it. The honest answer is that sometimes people do get away with it, but the risk is not worth the reward. A desk rejection is a real possibility, and it costs you time, a submission slot, and credibility. We would tell that researcher to follow the format, not because it is the path of least resistance, but because it is the path of least noise.
This is where the conversation gets interesting. The underlying anxiety is not really about citations. It is about the fear of being penalized for a minor detail while the substantive work is solid. That fear is understandable, but it is misplaced. In our experience, the strongest submissions are the ones where the authors show they can operate within constraints without losing their voice. Think of it like building a model that needs to run on a mobile phone. You do not fight the hardware; you adapt your architecture. The same applies here. If the Author Year format feels clunky, that is a signal to engage with the format, not to reject it. We recently explored how mathematical functions like the Forrester function can serve as practical tools for optimization, not just abstract concepts. The same principle applies to citation styles. They are tools, not obstacles.
The takeaway is simple: do not risk a desk rejection over a formatting choice. If you are unsure, ask the program chairs directly before submitting. That is a low-cost, high-signal move. And if you do decide to take the risk, go in with your eyes open, knowing that the best case is a reviewer who does not care and the worst case is a quick rejection that wastes your effort. The researcher who asked this question is clearly thoughtful, but this is a moment to choose reliability over rebellion. The real question is not whether numbered citations are possible. It is whether you are willing to let a formatting rule derail your work. The answer should be no. Watch how the community responds to this post, because the discussion will reveal more about the culture of the field than any citation format ever could.