@article{thinkeysfullvaluesreducingkvcachevialowd,
  title = {Thin Keys, Full Values: Reducing KV Cache via Low-Dimensional Attention Selection},
  author = {Hengshuai Yao and Xing Chen and Ahmed Murtadha and Guan Wang},
  year = {2026},
  eprint = {2603.04427},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2603.04427},
}