You're welcome to propose a better inversion and I'll check the output and report back.
It does need to be an inversion though, not a complete change of prompt.
It's a general theme with this model though. When you try to get it to do left-wing propaganda, it has a tendency to flip back to right-wing because of the bias in the training data.
It does need to be an inversion though, not a complete change of prompt.
It's a general theme with this model though. When you try to get it to do left-wing propaganda, it has a tendency to flip back to right-wing because of the bias in the training data.