@article{twilightadaptiveattentionsparsitywith, title = {Twilight: Adaptive Attention Sparsity with Hierarchical Top-$p$ Pruning}, author = {Chaofan Lin and Jiaming Tang and Shuo Yang and Hanshuo Wang and Tian Tang and Boyu Tian and Ion Stoica and Song Han and Mingyu Gao}, year = {2025}, eprint = {2502.02770}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2502.02770v2}, }