AAAI 2026technical0 citations
D²Pruner: Debiased Importance and Structural Diversity for MLLM Token Pruning
Evelyn Zhang, Fufu Yu, Aoqi Wu, Zichen Wen, Ke Yan, Shouhong Ding, Biqing Qi, Linfeng Zhang
Abstract
Processing long visual token sequences poses a significant computational burden on Multimodal Large Language Models (MLLMs). While token pruning offers a path to acceleration, we find that current methods, while adequate for general understanding, catastrophically fail on fine-grained localization tasks. We attribute this failure to the inherent flaws of the two prevailing strategies: importance-based methods suffer from a strong positional bias, an inherent model artifact that distracts from semantic content, while diversity-based methods exhibit structural blindness, disregarding the user
BibTeX
@inproceedings{aaai2026_dprunerdebiasedi,
title = {D²Pruner: Debiased Importance and Structural Diversity for MLLM Token Pruning},
author = {Evelyn Zhang and Fufu Yu and Aoqi Wu and Zichen Wen and Ke Yan and Shouhong Ding and Biqing Qi and Linfeng Zhang},
booktitle = {AAAI 2026},
year = {2026}
}