@article{cosaacceleratinglongcontextinferenceviap,
  title = {CoSA: Accelerating Long-Context Inference via Proxy-Kernel Co-Designed Sparse Attention},
  author = {Yufei Xue and Lin Niu and Hong Liu and Siran Liu and Hanyong Shao and Wei Liu and Guanghua Yu and Jianchen Zhu and Jun Zhang},
  year = {2026},
  eprint = {2607.25291},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2607.25291},
}