How to eval the accuracy/quality of a LLM?

cloud712
New Contributor

extremely subjective (human responses on likert scale don't suffice and hard to quantify with one metric on accuracy). curious how others have approached this