FIX Pipeline to use metadata passed to transform_input in intermediate steps - #34201

On November 23, 2025, you closed my PR as AI spam without offering any technical review of the actual changes.

A few months later, you submitted PR #34201 and explicitly disclosed using AI for code generation and research. That PR was reviewed and merged.

So apparently AI-generated code is not the problem.

The problem seems to be who is submitting it.

If there is an objective standard that explains why my contribution was dismissed as spam while yours was given a proper review, I would genuinely like to know what that standard is..

👯

#32769
#34201

See our automated contribution policy.

As human maintainers, we are doing our best while being kind of DDoS-attacked by a lot of accounts opening what seems like automated PRs.

As a personal opinion at the time of writing, when I see a PR from a new contributor I do a quick vibe-check: I look at a profile activity, if the account has opened a number of PRs in separate repos in a short amount of time, I am not interested in reviewing the PR and I'd rather try to help a slow human who is genuinely interested to contributing to the project. I may not reply, I may close the PR directly, I may block the user in the worst case. All of this is very fuzzy and I am not saying I am good at having a consistent rule.

If you are genuinely interested in contributing to scikit-learn and you haven't done it already, please have a look at our contributing doc and in particular the section Issues for new contributors.

I would suggest you look at the GitHub's How to contribute to Open-Source guide, which gives a good generic overview about how open-source works.