Building trust in this model and managing expectations
#1
by
zzlukes
- opened
Hi I am impressed with this model overall, so congratulations on great work. However, I have a few questions in my kit bag that many LLMs have struggled with and here is an example:
Chat request: "How many times does the letter r appear in strawberry?"
- spoiler - this model doesn't provide the correct / expected response - however it does iterate through chat in a friendly way
Are there any recommended adjustments, at inference time, e.g. with prompting, or RAG, to guide the model so it performs?