All committers have signed the CLA.

Hi @nelsonauner !

Thank you for pointing out this issue! Good catch on the unstable sorting algorithm.
It turns out there are just too many sets of exact duplicates in this tutorial, leading to varying results.

I've made some changes to ensure a more robust and clear approach.

• In the tutorial, we display the lowest-scoring example (based on the near-duplicate score) and its associated near-duplicate sets. I don't see a reason for us to hard-code which examples to display.

• I also updated the relevant text to be agnostic to specific hardcoded IDs (690, 246, etc.).

Finally, I added tests at the end of the notebook to ensure that all displayed rows are identical (since we state they are exact duplicates).

The ideal situation would be if we were able to add more interactivity to the docs, like an interactive widget:

Unfortunately, this isn't so straightforward when we build the docs with these notebooks.