claim
Plain vector similarity search in RAG systems is difficult to tune because the ideal number of retrieved documents varies by question, and retrieving too many documents can increase noise and cost.

Authors

Sources

Referenced by nodes (1)