In the Transformer architecture, the primary mechanism that connects the encoder and decoder is called ____. This mechanism allows for parallelization and has improved the efficiency of training models compared to traditional methods.

In a Transformer the encoder and decoder talk to each other through a mechanism called attention, specifically the scaled dot‑product attention. Attention lets the model look at every part of the input sequence at once, so all words can be processed in parallel instead of one after another. This parallelism speeds up training and lets the model learn long‑range relationships more easily. For example, if the decoder wants to generate the word “cat” after “the big ___”, attention lets it weigh the word “big” and the earlier word “the” simultaneously, rather than waiting for each step. Because everything happens at the same time, Transformers train much faster than older sequential models.