For next fine-tune
#8
by Hypersniper - opened
On the next fine-tune, try to reduce the number of em "—" dashes the model puts out. I'm not sure what's going on, but it seems like every model overuses them, and it’s a telltale sign that the text was written by an LLM.