What is attention mechanism in classification models?

Attention allows models to focus on important parts of input data for classification.

Attention computes weighted importance scores over input tokens or features, improving performance in NLP and vision tasks. Transformers rely heavily on self-attention.