新生成范式：从数据出发，用不变分布采样提升生成质量
Conservative Flows: A New Paradigm of Generative Models
- 不从噪声开始，而是从数据态出发进行随机动力学采样
- 在瑞士卷、ImageNet-256、Oxford Flowers-102上均优于原始生成方法
- 兼容预训练流模型，可直接使用现有模型检查点
现代生成建模主要依赖从噪声先验到数据的传输过程。本文提出一种新范式：生成通过保持数据分布不变的离散随机动力学实现，初始化于数据支持的状态而非噪声。该框架可利用任意预训练流模型。我们设计了两种概率保全的采样机制——带梅特罗波利斯修正的校正朗之万动力学与预测-校正流，均可直接作用于现有模型检查点。在合成瑞士卷目标、ImageNet-256 和 Oxford Flowers-102 数据集上的验证表明，所提采样器在所有场景下均持续优于原始生成流程。
原文摘要 · Abstract (English)
Modern generative modeling is dominated by transport from a noise prior to data. We propose an alternative paradigm in which generation is performed by a discrete stochastic dynamics that leaves the data distribution invariant, initialized from data-supported states rather than from noise. The framework can utilize any pretrained flow model. We develop two probability-preserving sampling mechanisms, a corrected Langevin dynamics with a Metropolis adjustment and a predictor-corrector flow, that operate directly on existing checkpoints. We validate the framework on a synthetic Swiss-roll target, ImageNet-256 and Oxford Flowers-102, where our samplers consistently improve over the original generation procedures.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。