Embeddings
Also known as: Vector representations, Dense vectors
Dense numeric vectors that capture meaning; close vectors = similar text.
Dense vector representations of text (or other modalities) in a continuous space where semantic similarity is captured by geometric distance. Used as the backbone for retrieval, classification, and clustering in modern AI systems.
In practice
The substrate for RAG, semantic search, and clustering. Interviews test embedding choice (BGE vs OpenAI vs E5), dimension trade-offs, and cosine vs dot product.
How it compares
Embeddings are the vectors themselves; vector databases are the storage + ANN index for querying them at scale.
Comparisons that include Embeddings
Related topics
Questions that mention this term
Related terms
Retrieval-Augmented Generation (RAG)
Fetch relevant docs at inference time and stuff them into the prompt so the model can answer from real data.
Vector Database
A database tuned for fast approximate nearest-neighbor search over millions of high-dimensional vectors.
HNSW (Hierarchical Navigable Small World)
A graph-based ANN index with log-time approximate search, excellent recall/latency trade-off.
Embedding Model
The model that produces embeddings, typically a smaller encoder-only transformer.
Cosine Similarity
Compare two vectors by the cosine of their angle; the default similarity metric for embedding search.
Hybrid Search
Combine BM25 (keywords) + dense embeddings (meaning); better recall than either alone.