EVATok: Adaptive Length Video Tokenization for Efficient Visual Autoregressive Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xiong, Tianwei, Liew, Jun Hao, Huang, Zilong, Lin, Zhijie, Feng, Jiashi, Liu, Xihui
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917336010719232
author Xiong, Tianwei
Liew, Jun Hao
Huang, Zilong
Lin, Zhijie
Feng, Jiashi
Liu, Xihui
author_facet Xiong, Tianwei
Liew, Jun Hao
Huang, Zilong
Lin, Zhijie
Feng, Jiashi
Liu, Xihui
contents Autoregressive (AR) video generative models rely on video tokenizers that compress pixels into discrete token sequences. The length of these token sequences is crucial for balancing reconstruction quality against downstream generation computational cost. Traditional video tokenizers apply a uniform token assignment across temporal blocks of different videos, often wasting tokens on simple, static, or repetitive segments while underserving dynamic or complex ones. To address this inefficiency, we introduce $\textbf{EVATok}$, a framework to produce $\textbf{E}$fficient $\textbf{V}$ideo $\textbf{A}$daptive $\textbf{Tok}$enizers. Our framework estimates optimal token assignments for each video to achieve the best quality-cost trade-off, develops lightweight routers for fast prediction of these optimal assignments, and trains adaptive tokenizers that encode videos based on the assignments predicted by routers. We demonstrate that EVATok delivers substantial improvements in efficiency and overall quality for video reconstruction and downstream AR generation. Enhanced by our advanced training recipe that integrates video semantic encoders, EVATok achieves superior reconstruction and state-of-the-art class-to-video generation on UCF-101, with at least 24.4% savings in average token usage compared to the prior state-of-the-art LARP and our fixed-length baseline.
format Preprint
id arxiv_https___arxiv_org_abs_2603_12267
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle EVATok: Adaptive Length Video Tokenization for Efficient Visual Autoregressive Generation
Xiong, Tianwei
Liew, Jun Hao
Huang, Zilong
Lin, Zhijie
Feng, Jiashi
Liu, Xihui
Computer Vision and Pattern Recognition
Autoregressive (AR) video generative models rely on video tokenizers that compress pixels into discrete token sequences. The length of these token sequences is crucial for balancing reconstruction quality against downstream generation computational cost. Traditional video tokenizers apply a uniform token assignment across temporal blocks of different videos, often wasting tokens on simple, static, or repetitive segments while underserving dynamic or complex ones. To address this inefficiency, we introduce $\textbf{EVATok}$, a framework to produce $\textbf{E}$fficient $\textbf{V}$ideo $\textbf{A}$daptive $\textbf{Tok}$enizers. Our framework estimates optimal token assignments for each video to achieve the best quality-cost trade-off, develops lightweight routers for fast prediction of these optimal assignments, and trains adaptive tokenizers that encode videos based on the assignments predicted by routers. We demonstrate that EVATok delivers substantial improvements in efficiency and overall quality for video reconstruction and downstream AR generation. Enhanced by our advanced training recipe that integrates video semantic encoders, EVATok achieves superior reconstruction and state-of-the-art class-to-video generation on UCF-101, with at least 24.4% savings in average token usage compared to the prior state-of-the-art LARP and our fixed-length baseline.
title EVATok: Adaptive Length Video Tokenization for Efficient Visual Autoregressive Generation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.12267