Might want to reach out to the author of the above benchmark. I think having a working group of multiple distinct benchmark owners might help quickly evaluate new model developments