TY  - RPRT
TI  - Video-XL-2: Towards Very Long-Video Understanding Through Task-Aware KV Sparsification
AU  - Minghao Qin
AU  - Xiangrui Liu
AU  - Zhengyang Liang
AU  - Yan Shu
AU  - Huaying Yuan
AU  - Juenjie Zhou
AU  - Shitao Xiao
AU  - Bo Zhao
AU  - Zheng Liu
PY  - 2025
UR  - https://arxiv.org/abs/2506.19225
ID  - 2506.19225
ER  -