TY  - RPRT
TI  - DVD: Dynamic Vector Decoding for Efficient MLLM-based Perception
AU  - Jinghua Hou
AU  - Zhe Liu
AU  - Hengshuang Zhao
PY  - 2026
UR  - https://arxiv.org/abs/2610.12266
ID  - 2610.12266
ER  -