nice! didn’t think to run this generally it seems like models that have less post training devolve much more into repetition, which makes some sense.
nice! didn’t think to run this generally it seems like models that have less post training devolve much more into repetition, which makes some sense.