Numerical representations of text where each word or sentence is converted into a list of numbers that capture its meaning, allowing the model to compare semantic similarity.
Performance retention over long documents and conversations