Tag
This paper introduces a grammatically-guided sparse attention mechanism for Transformers, aiming to improve efficiency and interpretability by leveraging linguistic structure.