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

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 advance 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 that 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 a minimal FID increase from 3.40 to 5.43 and 4.11 to 7.58 on ImageNet-256, while achieving 8.0× and 238× speedup with VAR and LlamaGen models, respectively. 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× 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.