This is why I want to see an example – just one – of a successful fine-tune that added specific extra knowledge to a model Until I've seen that I can't say for certain that it can be done

→ View original post on X — @simonw