Mean Pooling

A method that compresses all token hidden states in a sequence into a single vector by computing a weighted average according to the attention mask. It is commonly used to generate sentence-level embedding representations from transformer model outputs.