Output in different writing style
#4
by
marinl
- opened
I cannot get the LLM to output text in a different writing style than, what I consider, formal. It this because of the dataset it was trained on?
What system message have you been using? With other dolphin models, if you put in a system message like "you are a rude asshole who screams and is annoying", you would get non-formal responses. Without touching system message, responses would look like normal gptslop.
I put in your line as an example:
This is the experience I have generally with this model so far.