A deep learning architecture based on the self-attention mechanism, which weighs the significance of each part of the input data differently.
The Transformer Architecture is the design behind the current generation of AI models, including most LLMs. Using self-attention, Transformers process entire sequences of text in parallel (unlike sequential RNNs), which lets them capture long-range dependencies and context.
See also Self-Attention, Large Language Model, Fine-Tuning and Model Inference.