Embedding

An embedding is a numerical vector that represents an item so useful relationships can be measured or learned.

Plan for this page

Understand one concept in plain language and connect it to nearby ideas.

You will leave with: Models create embeddings for words, passages, images, users, products, or other objects. Items used similarly in the training context may lie near one another, but distance and meaning depend on the model and objective.
Time: 3–8 min
Before you begin: No prerequisite. Keep one real example in mind.
Do this now: Read the explanation, then compare the example with the common misunderstanding.

Prepared next stepSee retrieval and grounding

EmbeddingAn embedding is a numerical vector that represents an item so useful relationships can be measured or learned.

Feature
Token
Retrieval-augmented generation

Plain explanation

Models create embeddings for words, passages, images, users, products, or other objects. Items used similarly in the training context may lie near one another, but distance and meaning depend on the model and objective.

Example

A search system embeds a question and document passages, then retrieves passages whose vectors are nearby.

Why it matters

Embeddings enable retrieval, clustering, recommendation, and model inputs. They can also encode data biases, expose information, or produce misleading similarity outside their tested purpose.

Common misunderstanding

Misconception: an embedding is a perfect map of meaning. Correction: it is a task-shaped representation whose distances need contextual evaluation.

Related concepts

Follow relationships, not only the alphabet

FeatureA feature is an input property or representation used by a model when producing an output.TokenA token is a unit into which a model’s input or output is divided for processing.Retrieval-augmented generationRetrieval-augmented generation, or RAG, retrieves relevant material and places it in context for a generative model to use.

See retrieval and grounding

Sources and scope

What supports this object

Machine Learning GlossaryGoogle for Developers

Defines embeddings as learned vector representations; geometric similarity reflects the training objective and data, not universal semantic truth.

Open source

Loading page content...

…