These are all censored models?
#3
by
Goldenblood56
- opened
Thank you The Bloke your awesome. I assume these are all censored models?
Yeah I'm not aware that any of the "I'm sorry.." type responses were removed from the datasets prior to the fine tuning.
So they're not specifically filtered, but they use datasets with 'censored' responses and that makes it into the fine tuned models as well.