The performance is great, but the censorship is ridiculous for me. I tried it as a backend for my game Guessix[1], but it would refuse for ridiculous reasons like "Cannot answer questions about copyrighted works like Harry Potter."
Iirc abliteration (ablation?) can be done without "training" and is pretty quick. It finds the individual weights related to the concept you want to ablate, and modifies those weights to "deactivate" them. Precision brain surgery, to anthropomorphize.
Do you mean like structured outputs? Unfortunately here the model is guided to explicitly tell you when you violate the rules and why, it can confuse it's system rules with the game rules and say you're not allowed to ask a question about copyrighted material etc.
1. https://guessix.com/