Attention

As seen in Transformer Architecture and Attention Is All You Need.

• Multi-Head Attention
• The Scaled Dot-Product Attention