Sharing code builds trust, yet protecting your work matters.

As NeurIPS 2026 approaches, the question of submitting code alongside research papers has sparked diverse opinions within the community.

3 min readMachine Learning

The decision to share code alongside a research paper is not a simple yes or no, and the hesitation expressed by the original poster is one we understand deeply. The tension between transparency and protection is real, especially in an era where the line between collaboration and appropriation can feel blurred. But let's be clear: the credibility of your work is not a bonus feature, it is the foundation. When you publish code, you are not just sharing lines of logic, you are inviting scrutiny, and that invitation is what separates a claim from a contribution. The worry about plagiarism is legitimate, but it should not paralyze you into obscurity.

Practically speaking, the choice comes down to what you are trying to prove and to whom. If your goal is to advance the field, your code is the evidence that your methodology is sound. Without it, reviewers and peers are left to take your word for it, and in a discipline built on reproducibility, that is a fragile position. The poster's concern about "current times" is not unfounded, but consider this: the researchers who will likely misuse your work are not the ones who need your code to do so. They will find a way regardless. What sharing does is create a record, a timestamped, verifiable trail that establishes priority and ownership. It is not a perfect defense, but it is a practical one.

That said, protecting your work does not mean locking it away. It means being intentional about how you share. Use licenses that explicitly state what others can and cannot do. Include clear documentation that attributes the original ideas to you. Share code that is complete but also annotated in a way that makes your thought process visible. This is not about hiding behind legalities, it is about setting boundaries that allow for constructive use while discouraging outright theft. The act of sharing becomes a form of leadership, not vulnerability, when it is paired with clear expectations.

The real takeaway here is not that you must choose between openness and safety, but that you can have both if you approach it with a strategy. Do not let the fear of the worst-case scenario stop you from engaging in the very practice that builds trust in your work. The researchers who will cite you, build on your ideas, and push your findings further are not the ones you should be worried about. Focus on them. Publish your code, protect it with intention, and let your contribution speak for itself. That is how you move from being a participant in the conversation to being a voice that shapes it.

From Machine Learning

I am curious what everyone will be doing. I myself am torn, on the one hand I understand it boosts a paper’s credibility but on the other hand I worry about plagiarism, especially during current times. Thoughts?

Read the original at Machine Learning