Comment on My finetuned models beat OpenAI's GPT-4parentComments−sheepscreek2yVery loosely, isn’t this what is happening inside most LLMs that have a “multi-head” mechanism?
Comments
Very loosely, isn’t this what is happening inside most LLMs that have a “multi-head” mechanism?