It looks like the last commit of exploitablerunner.py might have an issue. I'm not sure why entries would have been removed since the paper states that all unibench targets are run with ASAN and GDB. The logic is otherwise the same except for removing a few print statements as well.
Also curious why the ID field in data is 'None' for all the existing entries. The old version had both integer and 'None'. I see that it's not used when generating the gdb commands to run for each target. Perhaps this is an artifact of storing the target data records in a database or how the crash data was being organized to run gdb against each target/crash pairing.