OmniZip: Audio-Guided Dynamic Token Compression for Fast Omnimodal Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tao, Keda, Shao, Kele, Yu, Bohan, Wang, Weiqiang, liu, Jian, Wang, Huan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913044263600128
author Tao, Keda
Shao, Kele
Yu, Bohan
Wang, Weiqiang
liu, Jian
Wang, Huan
author_facet Tao, Keda
Shao, Kele
Yu, Bohan
Wang, Weiqiang
liu, Jian
Wang, Huan
contents Omnimodal large language models (OmniLLMs) have attracted increasing research attention of late towards unified audio-video understanding. However, the high computational cost of processing longer joint audio-video token sequences has become a key bottleneck. Existing token compression methods have not addressed the emerging need to jointly compress multimodal tokens. To bridge this gap, we present OmniZip, a training-free, audio-guided audio-visual token-compression framework that optimizes multimodal token representation and accelerates model inference. Specifically, OmniZip first identifies salient audio tokens, then computes an audio retention score for each time group to capture information density, thereby dynamically guiding video token pruning and preserving cues from audio anchors enhanced by cross-modal similarity. For each time window, OmniZip compresses the video tokens using an interleaved spatio-temporal scheme. Extensive results demonstrate the merits of OmniZip: it achieves a 3.42X inference speedup and a 1.4X memory reduction over other top-performing counterparts, while maintaining the performance of OmniLLMs without training.
format Preprint
id arxiv_https___arxiv_org_abs_2511_14582
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle OmniZip: Audio-Guided Dynamic Token Compression for Fast Omnimodal Large Language Models
Tao, Keda
Shao, Kele
Yu, Bohan
Wang, Weiqiang
liu, Jian
Wang, Huan
Computer Vision and Pattern Recognition
Omnimodal large language models (OmniLLMs) have attracted increasing research attention of late towards unified audio-video understanding. However, the high computational cost of processing longer joint audio-video token sequences has become a key bottleneck. Existing token compression methods have not addressed the emerging need to jointly compress multimodal tokens. To bridge this gap, we present OmniZip, a training-free, audio-guided audio-visual token-compression framework that optimizes multimodal token representation and accelerates model inference. Specifically, OmniZip first identifies salient audio tokens, then computes an audio retention score for each time group to capture information density, thereby dynamically guiding video token pruning and preserving cues from audio anchors enhanced by cross-modal similarity. For each time window, OmniZip compresses the video tokens using an interleaved spatio-temporal scheme. Extensive results demonstrate the merits of OmniZip: it achieves a 3.42X inference speedup and a 1.4X memory reduction over other top-performing counterparts, while maintaining the performance of OmniLLMs without training.
title OmniZip: Audio-Guided Dynamic Token Compression for Fast Omnimodal Large Language Models
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.14582