@article{balancekvkvcachecompressionthrough,
  title = {Streaming Attention Approximation via Discrepancy Theory},
  author = {Insu Han and Michael Kapralov and Ekaterina Kochetkova and Kshiteej Sheth and Amir Zandieh},
  year = {2025},
  eprint = {2502.07861},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2502.07861v2},
}