Authors:Yue Ding, Yiyan Ji, Jungang Li, Xuyang Liu, Xinlong Chen, Junfei Wu, Bozhou Li, Bohan Zeng, Yang Shi, Yushuo Guan, Yuanxing Zhang, Jiaheng Liu, Qiang Liu, Pengfei Wan, Liang Wang
Abstract:Omni-modal Large Language Models (Omni-LLMs) have demonstrated strong capabilities in audio-video understanding tasks. However, their reliance on long multimodal token sequences leads to substantial computational overhead. Despite this challenge, token compression methods designed for Omni-LLMs remain limited. To bridge this gap, we propose OmniSIFT (Omni-modal Spatio-temporal Informed Fine-grained Token compression), a modality-asymmetric token compression framework tailored for Omni-LLMs. Specifically, OmniSIFT adopts a two-stage compression strategy: (i) a spatio-temporal video pruning module that removes video redundancy arising from both intra-frame structure and inter-frame overlap, and (ii) a vision-guided audio selection module that filters audio tokens. The entire framework is optimized end-to-end via a differentiable straight-through estimator. Extensive experiments on five representative benchmarks demonstrate the efficacy and robustness of OmniSIFT. Notably, for Qwen2.5-Omni-7B, OmniSIFT introduces only 4.85M parameters while maintaining lower latency than training-free baselines such as OmniZip. With merely 25% of the original token context, OmniSIFT consistently outperforms all compression baselines and even surpasses the performance of the full-token model on several tasks.
| Comments: | [ICML 2026] Code Link: this https URL |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2602.04804 [cs.CL] |
| (or arXiv:2602.04804v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2602.04804 arXiv-issued DOI via DataCite |
Submission history
From: Yue Ding [view email]
[v1]
Wed, 4 Feb 2026 17:51:05 UTC (1,840 KB)
[v2]
Wed, 13 May 2026 17:11:48 UTC (1,843 KB)