TY  - RPRT
TI  - InfiniteVL: Synergizing Linear and Sparse Attention for Highly-Efficient, Unlimited-Input Vision-Language Models
AU  - Hongyuan Tao
AU  - Bencheng Liao
AU  - Shaoyu Chen
AU  - Haoran Yin
AU  - Qian Zhang
AU  - Wenyu Liu
AU  - Xinggang Wang
PY  - 2026
UR  - https://arxiv.org/abs/2512.08829
ID  - 2512.08829
ER  -