Distilled Decoding 2: One-step Sampling of Image Auto-regressive Models with Conditional Score Distillation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liu, Enshu, Chen, Qian, Ning, Xuefei, Yan, Shengen, Dai, Guohao, Lin, Zinan, Wang, Yu
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914111187582976
author Liu, Enshu
Chen, Qian
Ning, Xuefei
Yan, Shengen
Dai, Guohao
Lin, Zinan
Wang, Yu
author_facet Liu, Enshu
Chen, Qian
Ning, Xuefei
Yan, Shengen
Dai, Guohao
Lin, Zinan
Wang, Yu
contents Image Auto-regressive (AR) models have emerged as a powerful paradigm of visual generative models. Despite their promising performance, they suffer from slow generation speed due to the large number of sampling steps required. Although Distilled Decoding 1 (DD1) was recently proposed to enable few-step sampling for image AR models, it still incurs significant performance degradation in the one-step setting, and relies on a pre-defined mapping that limits its flexibility. In this work, we propose a new method, Distilled Decoding 2 (DD2), to further advances the feasibility of one-step sampling for image AR models. Unlike DD1, DD2 does not without rely on a pre-defined mapping. We view the original AR model as a teacher model which provides the ground truth conditional score in the latent embedding space at each token position. Based on this, we propose a novel \emph{conditional score distillation loss} to train a one-step generator. Specifically, we train a separate network to predict the conditional score of the generated distribution and apply score distillation at every token position conditioned on previous tokens. Experimental results show that DD2 enables one-step sampling for image AR models with an minimal FID increase from 3.40 to 5.43 on ImageNet-256. Compared to the strongest baseline DD1, DD2 reduces the gap between the one-step sampling and original AR model by 67%, with up to 12.3$\times$ training speed-up simultaneously. DD2 takes a significant step toward the goal of one-step AR generation, opening up new possibilities for fast and high-quality AR modeling. Code is available at https://github.com/imagination-research/Distilled-Decoding-2.
format Preprint
id arxiv_https___arxiv_org_abs_2510_21003
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Distilled Decoding 2: One-step Sampling of Image Auto-regressive Models with Conditional Score Distillation
Liu, Enshu
Chen, Qian
Ning, Xuefei
Yan, Shengen
Dai, Guohao
Lin, Zinan
Wang, Yu
Machine Learning
Image Auto-regressive (AR) models have emerged as a powerful paradigm of visual generative models. Despite their promising performance, they suffer from slow generation speed due to the large number of sampling steps required. Although Distilled Decoding 1 (DD1) was recently proposed to enable few-step sampling for image AR models, it still incurs significant performance degradation in the one-step setting, and relies on a pre-defined mapping that limits its flexibility. In this work, we propose a new method, Distilled Decoding 2 (DD2), to further advances the feasibility of one-step sampling for image AR models. Unlike DD1, DD2 does not without rely on a pre-defined mapping. We view the original AR model as a teacher model which provides the ground truth conditional score in the latent embedding space at each token position. Based on this, we propose a novel \emph{conditional score distillation loss} to train a one-step generator. Specifically, we train a separate network to predict the conditional score of the generated distribution and apply score distillation at every token position conditioned on previous tokens. Experimental results show that DD2 enables one-step sampling for image AR models with an minimal FID increase from 3.40 to 5.43 on ImageNet-256. Compared to the strongest baseline DD1, DD2 reduces the gap between the one-step sampling and original AR model by 67%, with up to 12.3$\times$ training speed-up simultaneously. DD2 takes a significant step toward the goal of one-step AR generation, opening up new possibilities for fast and high-quality AR modeling. Code is available at https://github.com/imagination-research/Distilled-Decoding-2.
title Distilled Decoding 2: One-step Sampling of Image Auto-regressive Models with Conditional Score Distillation
topic Machine Learning
url https://arxiv.org/abs/2510.21003