Skip to content

AI Foundations

Multi-head attention

An AI method that examines several different kinds of relationships between pieces of information at the same time.

Example

A transformer combines information from multiple attention heads.

Why people use it

It lets a transformer combine several learned ways of attending to information.

What you'll hear

“Different heads can focus on different relationships.”

What this means for you

Avoid assigning a simple meaning to a head without supporting analysis.

Can you control it?

Developer-only

The people building or running the AI choose this setup. An everyday user generally needs their help to change how this part works.

Common questions

Does each part have a neat human-readable job?
Not necessarily. The learned roles can overlap and may not correspond to simple ideas people can name.
Does adding more heads always help?
No. More heads add complexity and do not guarantee better performance.
Can heads learn overlapping behavior?
Yes. They do not necessarily divide the work into distinct, easily named roles.

Related terms

Still have questions?

Up to 500 characters.

Ask LATHIC about AI. Relevant glossary entries may be included.

Your question, the glossary entries it matches, and a rotating pseudonymous identifier go to Microsoft Azure’s OpenAI service through Vercel AI Gateway to generate an answer. Zero retention and no training are required of the provider, and LATHIC does not save your question or answer. Privacy Notice