Also, LLMs don’t “love” anything. They predict what the most likely next word in a sentence will be, based on their training data. So if they’re predicting em dashes it’s because they were exposed to them when they were trained on edited prose by real writers
add a skeleton here at some point
3 months ago