Filter, Correlate, Compress: Training-Free Token Reduction for MLLM Acceleration

Yuhang Han; Xuyang Liu; Zihan Zhang; Pengxiang Ding; Junjie Chen; Honggang Chen; Donglin Wang; Qingsen Yan

AAAI 2026technical0 citations

Filter, Correlate, Compress: Training-Free Token Reduction for MLLM Acceleration

Yuhang Han, Xuyang Liu, Zihan Zhang, Pengxiang Ding, Junjie Chen, Honggang Chen, Donglin Wang, Qingsen Yan

Abstract

The quadratic complexity of Multimodal Large Language Models (MLLMs) with respect to context length poses significant computational and memory challenges, hindering their real-world deployment. In the paper, we devise a

BibTeX

@inproceedings{aaai2026_filtercorrelatec,
  title = {Filter, Correlate, Compress: Training-Free Token Reduction for MLLM Acceleration},
  author = {Yuhang Han and Xuyang Liu and Zihan Zhang and Pengxiang Ding and Junjie Chen and Honggang Chen and Donglin Wang and Qingsen Yan and Siteng Huang},
  booktitle = {AAAI 2026},
  year = {2026}
}