Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Deng, Yuchen, Cai, Zidang, Zheng, Hai-Tao, Wang, Jie, Yang, Feidiao, Han, Yuxing
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2605.12056
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918496970997760
author Deng, Yuchen
Cai, Zidang
Zheng, Hai-Tao
Wang, Jie
Yang, Feidiao
Han, Yuxing
author_facet Deng, Yuchen
Cai, Zidang
Zheng, Hai-Tao
Wang, Jie
Yang, Feidiao
Han, Yuxing
contents Omnimodal large language models (Omni-LLMs) show strong capability in audio-video understanding, but their practical deployment remains limited by high inference cost of long video streams and dense audio sequences. Despite recent progress, existing compression methods for Omni-LLMs typically rely on fixed or native compression units, which can disrupt cross-modal correspondence and the complementary information required for audio-video reasoning, making it difficult to improve inference efficiency while stably preserving performance. To address this, we propose OmniRefine, a training-free two-stage framework for efficient audio-visual token compression in Omni-LLMs. First, Correspondence-Preserving Chunk Refinement refines native chunk boundaries into cross-modally aligned compression units through frame-audio similarity and dynamic programming. Second, Modality-Aware Cooperative Compression jointly compresses video and audio tokens within each refined unit to reduce redundancy while preserving critical evidence. Extensive experiments show that OmniRefine achieves a better efficiency-performance trade-off than strong baselines and maintains stable performance under lower compression ratios. On WorldSense, it still reaches 46.7% accuracy at a 44% token retention ratio, nearly matching the full-token baseline. The code and interface will be released to facilitate further research.
format Preprint
id arxiv_https___arxiv_org_abs_2605_12056
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle OmniRefine: Alignment-Aware Cooperative Compression for Efficient Omnimodal Large Language Models
Deng, Yuchen
Cai, Zidang
Zheng, Hai-Tao
Wang, Jie
Yang, Feidiao
Han, Yuxing
Artificial Intelligence
Omnimodal large language models (Omni-LLMs) show strong capability in audio-video understanding, but their practical deployment remains limited by high inference cost of long video streams and dense audio sequences. Despite recent progress, existing compression methods for Omni-LLMs typically rely on fixed or native compression units, which can disrupt cross-modal correspondence and the complementary information required for audio-video reasoning, making it difficult to improve inference efficiency while stably preserving performance. To address this, we propose OmniRefine, a training-free two-stage framework for efficient audio-visual token compression in Omni-LLMs. First, Correspondence-Preserving Chunk Refinement refines native chunk boundaries into cross-modally aligned compression units through frame-audio similarity and dynamic programming. Second, Modality-Aware Cooperative Compression jointly compresses video and audio tokens within each refined unit to reduce redundancy while preserving critical evidence. Extensive experiments show that OmniRefine achieves a better efficiency-performance trade-off than strong baselines and maintains stable performance under lower compression ratios. On WorldSense, it still reaches 46.7% accuracy at a 44% token retention ratio, nearly matching the full-token baseline. The code and interface will be released to facilitate further research.
title OmniRefine: Alignment-Aware Cooperative Compression for Efficient Omnimodal Large Language Models
topic Artificial Intelligence
url https://arxiv.org/abs/2605.12056