Open vitalyshalumov opened 7 months ago
Thank you, My problem is not on a test set - it is on a per query metric: I want to let the user know the quality of the answer by showing him the metrics:
[Faithfulness] [Answer relevancy] [Context recall] [Context precision] [Context relevancy] [Context entity recall]
I have some code in WIP for verifiers that include such things, but not done. RAGAS is ok, but it's a bit loose compared to specific checking of actual specific faqs like done here: https://github.com/h2oai/enterprise-h2ogpte/tree/main/rag_benchmark