AOT: Token Reduction via Local and Global Contexts Optimization for Efficient Video Large Language Models
If you use AOT in academic or industrial research, please cite:
@article{li2026token,
title={Token Reduction via Local and Global Contexts Optimization for Efficient Video Large Language Models},
author={Li, Jinlong and Jiang, Liyuan and Zhang, Haonan and Sebe, Nicu},
journal={arXiv preprint arXiv:2603.01400},
year={2026}
}- Code: MIT License (see
LICENSE). - Model weights: Adobe Research License (see
LICENSE-WEIGHTS). The model weights are not covered by the MIT License.

