@article{improvingtransformerswithdynamically, title = {Improving Transformers with Dynamically Composable Multi-Head Attention}, author = {Da Xiao and Qingye Meng and Shengping Li and Xingyuan Yuan}, year = {2024}, eprint = {2405.08553}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2405.08553v2}, }