2

Many text model system prompts begin with "You are an AI assistant. You are helpful. You …. [user input] [model output]", but looking at what might have been in the training material, I would expect other perspectives to be more useful, such as

  • "I am an expert. Ask me your questions. [user input] [model output]"
  • "The following is a transcript of an expert explaining a topic to an interested user: [user input] [model output]"
  • "Top list of the best answers to [user input]: 1) [model output]"

Or similar prompts that (possibly) match high quality answers in a large uncurated corpus of crawled text.

The "You are ... You do ..." prompts seem to imply that they are not processed in the way one would expect a transformer text generation model to work, but would require some sort of self-awareness. Why and how do they work?

allo
  • 310
  • 1
  • 9
  • Transformers have attention. It's not guaranteed to take "you are" vs "I am" into account. I googled and found a study on the variance the role (expert, helpful assistant, etc) produces on the model output so presumably that is. I'd be interested to a see a definitive answer other than "although unnecessary, the engineers decided to personify their model in system prompts." – foreverska Dec 21 '23 at 21:53

0 Answers0