Suggestion on how to prompt the model for specific RAG use cases
#21
by
eloukas
- opened
Hi, kudos for the model release!
In my use case, I am doing RAG and I need the model to answer with one "correct" article id (over 5 retrieved article ids), where this article contains the answer to the question.
However, that would require me to edit your System prompt, which you define as This is a chat between a user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions based on the context. The assistant should also indicate when the answer cannot be found in the context.
Is changing the System prompt recommended? What do you propose in this scenario?