D$^{3}$ToM: Decider-Guided Dynamic Token Merging for Accelerating Diffusion MLLMs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chang, Shuochen, Zhang, Xiaofeng, Liu, Qingyang, Niu, Li
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909905445715968
author Chang, Shuochen
Zhang, Xiaofeng
Liu, Qingyang
Niu, Li
author_facet Chang, Shuochen
Zhang, Xiaofeng
Liu, Qingyang
Niu, Li
contents Diffusion-based multimodal large language models (Diffusion MLLMs) have recently demonstrated impressive non-autoregressive generative capabilities across vision-and-language tasks. However, Diffusion MLLMs exhibit substantially slower inference than autoregressive models: Each denoising step employs full bidirectional self-attention over the entire sequence, resulting in cubic decoding complexity that becomes computationally impractical with thousands of visual tokens. To address this challenge, we propose D$^{3}$ToM, a Decider-guided dynamic token merging method that dynamically merges redundant visual tokens at different denoising steps to accelerate inference in Diffusion MLLMs. At each denoising step, D$^{3}$ToM uses decider tokens-the tokens generated in the previous denoising step-to build an importance map over all visual tokens. Then it maintains a proportion of the most salient tokens and merges the remainder through similarity-based aggregation. This plug-and-play module integrates into a single transformer layer, physically shortening the visual token sequence for all subsequent layers without altering model parameters. Moreover, D$^{3}$ToM employs a merge ratio that dynamically varies with each denoising step, aligns with the native decoding process of Diffusion MLLMs, achieving superior performance under equivalent computational budgets. Extensive experiments show that D$^{3}$ToM accelerates inference while preserving competitive performance. The code is released at https://github.com/bcmi/D3ToM-Diffusion-MLLM.
format Preprint
id arxiv_https___arxiv_org_abs_2511_12280
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle D$^{3}$ToM: Decider-Guided Dynamic Token Merging for Accelerating Diffusion MLLMs
Chang, Shuochen
Zhang, Xiaofeng
Liu, Qingyang
Niu, Li
Computer Vision and Pattern Recognition
Computation and Language
Machine Learning
Diffusion-based multimodal large language models (Diffusion MLLMs) have recently demonstrated impressive non-autoregressive generative capabilities across vision-and-language tasks. However, Diffusion MLLMs exhibit substantially slower inference than autoregressive models: Each denoising step employs full bidirectional self-attention over the entire sequence, resulting in cubic decoding complexity that becomes computationally impractical with thousands of visual tokens. To address this challenge, we propose D$^{3}$ToM, a Decider-guided dynamic token merging method that dynamically merges redundant visual tokens at different denoising steps to accelerate inference in Diffusion MLLMs. At each denoising step, D$^{3}$ToM uses decider tokens-the tokens generated in the previous denoising step-to build an importance map over all visual tokens. Then it maintains a proportion of the most salient tokens and merges the remainder through similarity-based aggregation. This plug-and-play module integrates into a single transformer layer, physically shortening the visual token sequence for all subsequent layers without altering model parameters. Moreover, D$^{3}$ToM employs a merge ratio that dynamically varies with each denoising step, aligns with the native decoding process of Diffusion MLLMs, achieving superior performance under equivalent computational budgets. Extensive experiments show that D$^{3}$ToM accelerates inference while preserving competitive performance. The code is released at https://github.com/bcmi/D3ToM-Diffusion-MLLM.
title D$^{3}$ToM: Decider-Guided Dynamic Token Merging for Accelerating Diffusion MLLMs
topic Computer Vision and Pattern Recognition
Computation and Language
Machine Learning
url https://arxiv.org/abs/2511.12280