The benchmark programs used in our paper can be found on GitHub.
They are in the directoryexamples/pldi18.bm if you checked out the whole artifact from Github.
We don't include the .spec for most safe benchmarks as CUBA anyway
computes the full reachability for them.