"Rich" answers don't hallucinate. Almost.
RAG2Rich defines and computes the "Rich" score of answers based on the context relevance (C ), answer relevance (A), and groundedness (G) measures provided by TruLens. The Rich score can be largely thought of as indicative "richness" of answers, on average. Specifically, the Rich metric is computed as:
where X = (G, C, A), P is a set of parameters, and W is a weight vector.
The currently used RAG parameters are:
- chunk size = 512
- chunk overlap = 75
- top-k = 4
- Cohere re-ranker top-n = 3