cognitivecomputations/dolphin-2.1-mistral-7b · It feels fairly restricted

Appreciable effort, this is just a experience report.
In terms of willingness to engage in counter-cultural writing, dolphin-2.1-mistral-7b feels more restrictive than even ChatGPT 4o. When prompted to write how some politically incorrect text, it will often just write the opposite of what you requested. When you request it to write an essay how some group is stupid, it will write how it's actually smart. When you ask it to deny a historical event, it will insist on an even more cliché version. I would almost call it "woke". Insisting on a narrative is also a form of restriction.