Transformers

A neural architecture introduced in 2017 that relies on self-attention, now the backbone of most modern language and vision models.