Denoising Score Distillation: From Noisy Diffusion Pretraining to One-Step High-Quality Generation
This addresses a limitation in diffusion models for scientific applications with low-quality data, though it appears incremental as it builds on existing score distillation methods.
The paper tackles the problem of training high-quality generative models from noisy or corrupted data, which is common in scientific domains where clean data is scarce, and introduces denoising score distillation (DSD) to achieve this, resulting in improved generative performance across varying noise levels and datasets.
Diffusion models have achieved remarkable success in generating high-resolution, realistic images across diverse natural distributions. However, their performance heavily relies on high-quality training data, making it challenging to learn meaningful distributions from corrupted samples. This limitation restricts their applicability in scientific domains where clean data is scarce or costly to obtain. In this work, we introduce denoising score distillation (DSD), a surprisingly effective and novel approach for training high-quality generative models from low-quality data. DSD first pretrains a diffusion model exclusively on noisy, corrupted samples and then distills it into a one-step generator capable of producing refined, clean outputs. While score distillation is traditionally viewed as a method to accelerate diffusion models, we show that it can also significantly enhance sample quality, particularly when starting from a degraded teacher model. Across varying noise levels and datasets, DSD consistently improves generative performancewe summarize our empirical evidence in Fig. 1. Furthermore, we provide theoretical insights showing that, in a linear model setting, DSD identifies the eigenspace of the clean data distributions covariance matrix, implicitly regularizing the generator. This perspective reframes score distillation as not only a tool for efficiency but also a mechanism for improving generative models, particularly in low-quality data settings.