Leonie (@helloiamleonie)

2024-06-11 | ❤️ 595 | 🔁 137


“RAG is only as good as the retrieved documents’ relevance.”

But how do you quantify the relevance of a piece of context to the user input? And how can you improve the relevance?

In my latest article, we discuss different metrics related to retrieval:

  • Why similarity doesn’t equal relevancy (cosine similarity, dot product, etc.)
  • Common evaluation metrics for search systems (Precision, recall, MRR, NDCG, etc.)
  • How context relevancy is calculated in frameworks such as @ragas_io, @TruLensML, and DeepEval

📌 Read more on @TDataScience: https://t.co/qwzbFJpP0D

미디어

photo


Tags

domain-xr domain-dev-tools