[zie1ony]

They are not, and that's the whole point of doing this research. If we can build good benchmark, models developers would have nice goal.