Transformer
Also known as: transformer architecture
The transformer is the architecture behind essentially all modern LLMs. Before it, models read text strictly left-to-right, one token at a time. The transformer reads a whole sequence at once and lets each token attend to every other.
That “look everywhere at once” design is what made long-range understanding possible — connecting a word on line 1 to a word on line 40.
Key idea: attention (each token computes how much to focus on every other token).
Why the name: it *transforms* an input sequence into an output sequence — text to text, text to image, etc. If you remember one architecture, make it this one.