@article{cui-etal-2025-att-diag, author={Cui, Yiming and Yao, Xin and Wang, Shijin and Hu, Guoping}, journal={IEEE Signal Processing Letters}, title={You Might Not Need Attention Diagonals}, year={2025}, volume={32}, number={}, pages={3435-3439}, keywords={Training;Computational modeling;Transformers;Natural language processing;Convergence;Optimization;Mathematical models;Encyclopedias;Tuning;Semantics;Artificial intelligence;large language models;natural language processing}, doi={10.1109/LSP.2025.3601497} }