Classification Head
A small output layer that maps the model’s understanding to a choice among fixed categories.
Think of it like
Like the final judge at a science fair who takes all the deliberation and stamps one ribbon: first, second, or third.
Example
A spam filter puts a classification head on a text encoder to output “spam” or “not spam.”
How it actually works
It takes a pooled representation of the input and projects it to one score per class, with the highest score winning. You can bolt it onto a frozen or fine-tuned backbone to specialize a general model for a specific labeling task. It’s the standard way to turn an understanding model into a decision-maker.
For product teams
The cheap add-on that turns a language model into a categorizer.
For engineers
A projection from a pooled representation to per-class logits for a fixed label set.
Related
- Output Head — The general concept it’s a version of.
- Pooling — How the input gets summarized for it.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome