@misc{indiciae07457e62e31a,
  title = {DVD: Dynamic Vector Decoding for Efficient MLLM-based Perception},
  author = {Jinghua Hou and Zhe Liu and Hengshuang Zhao},
  year = {2026},
  url = {https://arxiv.org/abs/2610.12266},
  note = {Source identifier: 2610.12266}
}