T5
Also called Text-to-Text Transfer Transformer
A model that treats every task — translation, summarizing, answering — as turning one string of text into another.
Think of it like
Like a universal adapter that insists everything, from questions to translations, plugs into the same text-in, text-out socket.
Example
To classify a review, you’d feed T5 “sentiment: this movie was great” and it outputs the word “positive.”
How it actually works
T5 uses the full encoder–decoder transformer and reframes every NLP task as text-to-text, so one model and one training setup handle them all. That uniformity made it a clean research workhorse. Decoder-only chat models later showed you could get similar generality without the encoder, but the “everything is text-to-text” framing stuck around.
For product teams
An elegant older approach; most new work uses decoder-only models instead.
For engineers
An encoder–decoder transformer casting all tasks in a unified text-to-text format.
Related
- Encoder-Decoder — The two-part structure it uses.
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