G$^2$TR: Generation-Guided Visual Token Reduction for Separate-Encoder Unified Multimodal Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Junxian, Liu, Kai, Ding, Zizhong, Wang, Zhixin, Chen, Zhikai, Pei, Renjing, Zhang, Yulun
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909045597667328
author Li, Junxian
Liu, Kai
Ding, Zizhong
Wang, Zhixin
Chen, Zhikai
Pei, Renjing
Zhang, Yulun
author_facet Li, Junxian
Liu, Kai
Ding, Zizhong
Wang, Zhixin
Chen, Zhikai
Pei, Renjing
Zhang, Yulun
contents The development of separate-encoder Unified multimodal models (UMMs) comes with a rapidly growing inference cost due to dense visual token processing. In this paper, we focus on understanding-side visual token reduction for improving the efficiency of separate-encoder UMMs. While this topic has been widely studied for MLLMs, existing methods typically rely on attention scores, text-image similarity and so on, implicitly assuming that the final objective is discriminative reasoning. This assumption does not hold for UMMs, where understanding-side visual tokens must also preserve the model's capabilities for editing images. We propose G$^2$TR, a generation-guided visual token reduction framework for separate-encoder UMMs. Our key insight is that the generation branch provides a task-agnostic signal for identifying understanding-side visual tokens that are not only semantically relevant but also important for latent-space image reconstruction and generation. G$^2$TR estimates token importance from consistency with VAE latent, performs balanced token selection, and merges redundant tokens into retained representatives to reduce information loss. The method is training-free, plug-and-play, and applied only after the understanding encoding stage, making it compatible with existing UMM inference pipelines. Experiments on image understanding and editing benchmarks show that G$^2$TR substantially reduces visual tokens and prefill computation by 1.94x while maintaining both reasoning accuracy and editing quality, outperforming baselines on almost all benchmarks. Code is at: https://github.com/lijunxian111/G2TR.
format Preprint
id arxiv_https___arxiv_org_abs_2605_12309
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle G$^2$TR: Generation-Guided Visual Token Reduction for Separate-Encoder Unified Multimodal Models
Li, Junxian
Liu, Kai
Ding, Zizhong
Wang, Zhixin
Chen, Zhikai
Pei, Renjing
Zhang, Yulun
Computer Vision and Pattern Recognition
The development of separate-encoder Unified multimodal models (UMMs) comes with a rapidly growing inference cost due to dense visual token processing. In this paper, we focus on understanding-side visual token reduction for improving the efficiency of separate-encoder UMMs. While this topic has been widely studied for MLLMs, existing methods typically rely on attention scores, text-image similarity and so on, implicitly assuming that the final objective is discriminative reasoning. This assumption does not hold for UMMs, where understanding-side visual tokens must also preserve the model's capabilities for editing images. We propose G$^2$TR, a generation-guided visual token reduction framework for separate-encoder UMMs. Our key insight is that the generation branch provides a task-agnostic signal for identifying understanding-side visual tokens that are not only semantically relevant but also important for latent-space image reconstruction and generation. G$^2$TR estimates token importance from consistency with VAE latent, performs balanced token selection, and merges redundant tokens into retained representatives to reduce information loss. The method is training-free, plug-and-play, and applied only after the understanding encoding stage, making it compatible with existing UMM inference pipelines. Experiments on image understanding and editing benchmarks show that G$^2$TR substantially reduces visual tokens and prefill computation by 1.94x while maintaining both reasoning accuracy and editing quality, outperforming baselines on almost all benchmarks. Code is at: https://github.com/lijunxian111/G2TR.
title G$^2$TR: Generation-Guided Visual Token Reduction for Separate-Encoder Unified Multimodal Models
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2605.12309