Out of Sight, Still in Mind: Token Compression for Omni-LLMs

2026-07-23

AI Summary

This work addresses the high input token cost and redundancy introduced by visual modalities in Omni-LLM inference by proposing ReMo, a novel framework that achieves training-free visual token compression for the first time. ReMo identifies cross-modal redundancy between vision and audio through embedding alignment, retaining only visual information that cannot be explained by other modalities. It further replaces object-level visual tokens with semantically equivalent yet compact textual descriptions. Evaluated on Qwen2.5-Omni at two scales, ReMo reduces input tokens by 54% on average while achieving 101.2%–101.3% of the original model’s accuracy across five audio-visual benchmarks, significantly enhancing inference efficiency without compromising performance.

Abstract

The goal of this paper is to reduce the input token cost of Omni-modal large language models (Omni-LLMs) at inference time. Omni-LLMs reason jointly over audio, video and text, but the cost of the three streams is highly unbalanced: visual tokens account for the vast majority of the input, and are highly redundant. In this paper, we propose ReMo, a training-free framework that compresses visual tokens by redistributing their information across modalities: a visual token is kept only if its information appears nowhere else. ReMo achieves this in two ways: (i) it aligns audio and video in a common embedding space, and removes visual tokens already explained by the audio or by other visual tokens; and (ii) it replaces object-level visual tokens with compact text proxies, short descriptions of each object and its location, conveying the same content in far fewer tokens. On Qwen2.5-Omni at two model scales, ReMo removes 54% of the input tokens with no loss in accuracy. Indeed, it slightly exceeds the full-token model, reaching 101.2% and 101.3% of its average accuracy over five audio-visual benchmarks.

Problem

• Omni-LLMs
• token compression
• visual redundancy
• multimodal inference
• input cost

Innovation

• token compression
• Omni-LLMs
• modality alignment
• text proxy
• training-free