Definition
A RAG pipeline failure where the retriever identifies semantically similar but factually irrelevant noise (low precision) or fails to surface the necessary ground-truth data (low recall), often requiring a trade-off between retrieval depth and system latency.
Failing to find the right data versus the LLM hallucinating despite having the right data.
"A librarian providing a stack of cookbooks to a patron who specifically asked for a car repair manual because both books use the word 'oil'."
- Recall(Performance Metric)
- Vector Database(Infrastructure Component)
- Semantic Search(Prerequisite)
- Hallucination(Downstream Consequence)
Conceptual Overview
A RAG pipeline failure where the retriever identifies semantically similar but factually irrelevant noise (low precision) or fails to surface the necessary ground-truth data (low recall), often requiring a trade-off between retrieval depth and system latency.
Disambiguation
Failing to find the right data versus the LLM hallucinating despite having the right data.
Visual Analog
A librarian providing a stack of cookbooks to a patron who specifically asked for a car repair manual because both books use the word 'oil'.