Question about sampling test cases

[Felix-Goosens]

Hello,
I am working on a research project about test case clustering.
To compare my results I want to use Igor on my dataset.
However, I have a question on how Igor should be used.

In the Igor paper, it was mentioned that the PoCs were sampled using their stack hash before they were used on IgorFuzz.
My question regards the following sentence from the paper: "If there are many PoCs mapping to the same unique stack hash, we only process the 50 most diverse PoCs".
However, it is not clear to me what metric was used to determine the diversity of the PoCs and how it was applied.
Do you know how I should sample the test cases when using Igor?

I asked the first author of the paper and was redirected here.
So, I would appreciate any clarification so I can properly use Igor and compare my results.