In a RAG architecture, what do embeddings enable — the core mechanism that retrieves relevant documents for the prompt?
A.Compressing documents to smaller file sizes
Embeddings aren't compression — they're semantic representations.
B.Semantic similarity search — matching meaning rather than keywords
Meaning-based retrieval = embedding similarity. Correct.
C.Encrypting documents for compliance
Embeddings aren't encryption.
D.Translating documents to English
Translation is a different capability entirely.
Why this answer
Embeddings map text to dense vectors where semantic similarity = geometric proximity. RAG retrieves docs whose embeddings are closest to the query embedding — the 'find relevant knowledge' step before generation. Core exam concept.
Correct answer: B · Last reviewed 2026-11-05
Official AWS references
Practice this interactively
Answer without seeing the solution first in the Generative AI Concepts practice session.
Practice Generative AI Concepts