No idea. The post about waiting for full peer review is a good one. I don't really do this kind of work anymore and don't have time (or computation really) to try to recreate their alignments, so I couldn't really tell you anymore than what the paper says.
I don't think the suggestion would be to not try and control behavior. It's still pretty bad.
I don't think the suggestion would be to not try and control behavior. It's still pretty bad.