@article{transformersprovablylearndirectedacyclic, title = {Transformers Provably Learn Directed Acyclic Graphs via Kernel-Guided Mutual Information}, author = {Yuan Cheng and Yu Huang and Zhe Xiong and Yingbin Liang and Vincent Y. F. Tan}, year = {2025}, eprint = {2510.25542}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2510.25542}, }