# What is Adapters? Definition and meaning

Adds compact bottleneck layers between the frozen layers of a pre-trained model and trains only those.

## Definition

**Adapters**: Adds compact bottleneck layers between the frozen layers of a pre-trained model and trains only those.

In short: Small trainable modules inserted into a frozen model.

## When it is used

Parameter-efficient multi-task adaptation.

## Learn more

The full guide to **[Adapters](/tools/adapters)** explains how it works, when to use it and how it compares with the alternatives, with examples. The [glossary](/glossary) lists every term on the site from A to Z.

Previous term: [Active learning](/glossary/active-learning) · Next term: [Advanced RAG](/glossary/advanced-rag)
