Implement Generative Quantum Advantage Tutorial

[MichaelBroughton]

We are pretty excited about the results here: https://arxiv.org/pdf/2509.09033

Would be good to implement a new tutorial for the website showcasing some of these results! Perhaps we could show off the IDQNN circuits and talk about the hardness results that go along with it.

[rosspeili]

What would be the desired context and flow of the tutorial @MichaelBroughton?
@mhucka could I take the time to digest the paper in its depth and suggest a tutorial draft similar to existing tutorials on the site, or a notebook? Thanks.

[mhucka]

@rosspeili If you'd like, yes.

I'm not completely clear if by "take the time to digest the paper" you mean that we shouldn't assign it to you yet. Either way is fine; just let us know (and no rush).

[rosspeili]

I am happy to take this on, if there is no time pressure, but I am uncertain how to approach this. Are we looking for a page like this? https://www.tensorflow.org/quantum/tutorials/qcnn and if yes, where should I submit this? is there a dedicated repo for the website? just an html page or js, notebook? Any pointers would be appraciated.

[mhucka]

@rosspeili Yes, that's the kind of tutorial being sought. The pages on tensorflow.org/quantum are actually Jupyter/Colab notebook documents; the sources to all the tutorials are in docs/tutorials/. (The file names are admittedly a bit cryptic, but if you click on the files, GitHub's interface should display the notebook in an embedded view.) Those notebook files are processed by an internal (to Google) system to generate the web pages on the site. The process runs nightly. To add a new notebook involves simply making a PR to add the .ipynb file and also add an entry to the file docs/_book.yaml under the section named "Guide & Tutorials".

The full tensorflow.org site-generation system is unfortunately not open source, so it's not possible to preview the full site pages externally. Thankfully, very little is done on .ipynb files besides converting them to HTML, so just making a Jupyter or Colab notebook look like the existing ones should result in more or less "what you see is what you get" on the final site. One of the TFQ project maintainers can preview the full thing using internal (to Google) tooling as part of the PR review.

Last note: there is an old unfinished PR for an example of quantum advantage (#649). This is not the same thing as generative quantum advantage (despite the confusingly similar names), but it might be worth looking at Michael Broughton's review comments, in case it gives hints about some things to do or not do.

[rosspeili]

Thanks a lot for the detailed direction @mhucka, this is clear.

[rosspeili]

@mhucka @MichaelBroughton let me know how this looks, any feedback or comments appreciated.