发表机构
University of Houston; The University of Texas at Arlington（休斯顿大学; 德克萨斯大学阿灵顿分校）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对通用大模型令牌压缩的偏向音频分配问题，提出无需训练的Macer压缩器，先分配音视频预算再按模态排序，降本同时在多基准上保持性能，在OmniVinci-9B上提升达12.9个点。
AI 中文摘要
通用大模型（OmniLLMs）中的令牌压缩通常被表述为单一的显著性排序问题：为每个多模态令牌打分，保留前K个。我们认为这种抽象定义不当。相同的注意力分数同时决定两件事：每个模态获得多少保留容量，以及模态内保留哪些令牌。因此，共享的前K规则继承了偏向音频的分配先验，在视频令牌有竞争机会之前，就将保留容量分配给音频。我们提出Macer，一种无需训练的压缩器，它首先为音频和视频分配明确的预算，然后在特定模态的浅层内执行分配归一化排序。Macer显著降低了令牌成本，同时在音频基础、音视频联合、视觉主导和视频中心基准上保持了准确性。在25%的保留率下，Macer在Qwen2.5-Omni-7B上保留了全令牌性能的98.7%，在Qwen2.5-Omni-3B上保留了97.3%。在Qwen2.5-Omni-7B上，该25%设置在45%保留率时达到了OmniZip级别的性能，同时使用更低的浮点运算量（FLOPs）。在OmniVinci-9B上，相同的“先分配后排序”原则比共享前K排序提升了多达12.9个点。
英文摘要
Token compression in OmniLLMs is typically posed as a single saliency-ranking problem: score each multimodal token, keep the top-K. We argue this abstraction is mis-specified. The same attention score simultaneously decides two things: how much retained capacity each modality receives, and which tokens within a modality are kept. A shared top-K rule therefore inherits this audio-favoring allocation prior, spending retained capacity on audio before video tokens have a chance to compete. We propose Macer, a training-free compressor that first assigns explicit audio and video budgets, then performs allocation-normalized ranking within each modality at modality-specific shallow layers. Macer significantly reduces token cost while preserving accuracy across audio-grounded, audio--video joint, visual-dominant, and video-centric benchmarks. At 25 % retention, Macer preserves 98.7 % of full-token performance on Qwen2.5-Omni-7B and 97.3 % on Qwen2.5-Omni-3B. On Qwen2.5-Omni-7B, this 25 % setting reaches OmniZip-level performance at 45 % retention while using lower FLOPs. On OmniVinci-9B, the same allocation-before-ranking principle improves over shared top-K ranking by up to 12.9 points.
