@article{topdowncompressionrevisitefficientvision,
  title = {Top-Down Compression: Revisit Efficient Vision Token Projection for Visual Instruction Tuning},
  author = {Bonan Li and ZiCheng Zhang and Songhua Liu and Weihao Yu and Xinchao Wang},
  year = {2025},
  eprint = {2505.11945},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2505.11945v1},
}