蒸馏解码2：基于条件分数蒸馏的图像自回归模型一步采样法

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

October 23, 2025

作者: Enshu Liu, Qian Chen, Xuefei Ning, Shengen Yan, Guohao Dai, Zinan Lin, Yu Wang

cs.AI

摘要

图像自回归模型已成为视觉生成模型的重要范式。尽管性能优异，但由于需要大量采样步骤，其生成速度较慢。虽然近期提出的蒸馏解码一代（DD1）实现了图像自回归模型的少步采样，但在单步采样场景下仍存在明显性能下降，且依赖预定义映射限制了灵活性。本研究提出新方法蒸馏解码二代（DD2），进一步推进图像自回归模型单步采样的可行性。与DD1不同，DD2无需依赖预定义映射。我们将原始自回归模型视为教师模型，其在潜在嵌入空间的每个标记位置提供真实条件分数。基于此，提出新颖的条件分数蒸馏损失函数来训练单步生成器：通过训练独立网络预测生成分布的条件分数，并在每个标记位置基于前序标记进行分数蒸馏。实验表明，DD2在ImageNet-256数据集上实现单步采样时，FID指标仅从3.40微增至5.43。相较于最强基线DD1，DD2将单步采样与原始自回归模型的性能差距缩小67%，同时训练速度最高提升12.3倍。DD2向单步自回归生成目标迈出重要一步，为快速高质量的AR建模开辟了新可能。代码已开源于https://github.com/imagination-research/Distilled-Decoding-2。

English

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 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 12.3times 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.