The neural network architecture underlying most modern large language models, built around a mechanism called self-attention.
Introduced in 2017, the transformer architecture made it possible to train much larger, more capable language models — it's the technical foundation behind the entire current generation of AI systems.
