

No, I make corrections and instructions that are a pain, but sometimes it works… Delete the message or just something like Note to Narrator: The character Joe is a kind, loving person towards animals, very calm and collected, a normal person who has a powerful love for Dogs. He would never act mean or angry towards any animal It will generally correct itself. Though sometimes it forgets again, and you just have to copy and paste it to the end of a message from above. lol… I also add those kinds of things to my prompt as well. But it’s hit or miss. It might not be the right way, but it mostly works for me.
All you can really do is go in and set some context and parameters ahead of time and gently guide and correct as you go. It works “pretty” well and I have pretty ,such learned the guidance through trail and error to get to where I need to go. Sometimes its constant reminders to avoid and regenerating a response several times to kind of guide. Its really IMO not better or worse than before as the previous LLM was overly kind and generous it took the some kinds of gental coaxing just from a different perspective.