Fast Inference of Visual Autoregressive Model with Adjacency-Adaptive Dynamical Draft Trees

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lei, Haodong, Wang, Hongsong, Geng, Xin, Wang, Liang, Zhou, Pan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912789761622016
author Lei, Haodong
Wang, Hongsong
Geng, Xin
Wang, Liang
Zhou, Pan
author_facet Lei, Haodong
Wang, Hongsong
Geng, Xin
Wang, Liang
Zhou, Pan
contents Autoregressive (AR) image models achieve diffusion-level quality but suffer from sequential inference, requiring approximately 2,000 steps for a 576x576 image. Speculative decoding with draft trees accelerates LLMs yet underperforms on visual AR models due to spatially varying token prediction difficulty. We identify a key obstacle in applying speculative decoding to visual AR models: inconsistent acceptance rates across draft trees due to varying prediction difficulties in different image regions. We propose Adjacency-Adaptive Dynamical Draft Trees (ADT-Tree), an adjacency-adaptive dynamic draft tree that dynamically adjusts draft tree depth and width by leveraging adjacent token states and prior acceptance rates. ADT-Tree initializes via horizontal adjacency, then refines depth/width via bisectional adaptation, yielding deeper trees in simple regions and wider trees in complex ones. The empirical evaluations on MS-COCO 2017 and PartiPrompts demonstrate that ADT-Tree achieves speedups of 3.13xand 3.05x, respectively. Moreover, it integrates seamlessly with relaxed sampling methods such as LANTERN, enabling further acceleration. Code is available at https://github.com/Haodong-Lei-Ray/ADT-Tree.
format Preprint
id arxiv_https___arxiv_org_abs_2512_21857
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Fast Inference of Visual Autoregressive Model with Adjacency-Adaptive Dynamical Draft Trees
Lei, Haodong
Wang, Hongsong
Geng, Xin
Wang, Liang
Zhou, Pan
Computer Vision and Pattern Recognition
Autoregressive (AR) image models achieve diffusion-level quality but suffer from sequential inference, requiring approximately 2,000 steps for a 576x576 image. Speculative decoding with draft trees accelerates LLMs yet underperforms on visual AR models due to spatially varying token prediction difficulty. We identify a key obstacle in applying speculative decoding to visual AR models: inconsistent acceptance rates across draft trees due to varying prediction difficulties in different image regions. We propose Adjacency-Adaptive Dynamical Draft Trees (ADT-Tree), an adjacency-adaptive dynamic draft tree that dynamically adjusts draft tree depth and width by leveraging adjacent token states and prior acceptance rates. ADT-Tree initializes via horizontal adjacency, then refines depth/width via bisectional adaptation, yielding deeper trees in simple regions and wider trees in complex ones. The empirical evaluations on MS-COCO 2017 and PartiPrompts demonstrate that ADT-Tree achieves speedups of 3.13xand 3.05x, respectively. Moreover, it integrates seamlessly with relaxed sampling methods such as LANTERN, enabling further acceleration. Code is available at https://github.com/Haodong-Lei-Ray/ADT-Tree.
title Fast Inference of Visual Autoregressive Model with Adjacency-Adaptive Dynamical Draft Trees
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.21857