不加噪点也能修正生成模型偏差，提升图像质量。
Rethinking Refinement: Correcting Generative Bias without Noise Injection
- 用两阶段流匹配方法，在潜空间和数据空间分别修正生成偏差。
- 在MNIST上从FID 3.95降到1.46，仅需一次函数求值。
- 无需扰动采样过程，适合追求高质量图像的生成应用。
生成模型（如扩散模型和流模型）在高维设置下常表现出系统性偏差，影响样本质量。本文重新审视精炼方法，证明有效的偏差校正可作为后处理步骤，无需注入噪声或多次重采样。提出基于流匹配的双阶段流精炼（BFR）框架，包含两个阶段：潜空间对齐以实现近似可逆生成器，以及使用轻量增强训练的数据空间精炼。与以往通过扰动采样动态的精炼器不同，BFR保持原始常微分方程（ODE）轨迹，对生成样本施加确定性修正。在MNIST、CIFAR-10和FFHQ（256x256分辨率）上的实验显示，一致提升保真度与覆盖度；尤其在MNIST上，从基础样本FID 3.95出发，潜空间精炼仅通过一次额外函数求值（1-NFE），达到状态最优的FID 1.46，同时维持样本多样性。
原文摘要 · Abstract (English)
Generative models, including diffusion and flow-based models, often exhibit systematic biases that degrade sample quality, particularly in high-dimensional settings. We revisit refinement methods and show that effective bias correction can be achieved as a post-hoc procedure, without noise injection or multi-step resampling of the sampling process. We propose a flow-matching-based \textbf{Bi-stage Flow Refinement (BFR)} framework with two refinement strategies operating at different stages: latent space alignment for approximately invertible generators and data space refinement trained with lightweight augmentations. Unlike previous refiners that perturb sampling dynamics, BFR preserves the original ODE trajectory and applies deterministic corrections to generated samples. Experiments on MNIST, CIFAR-10, and FFHQ at 256x256 resolution demonstrate consistent improvements in fidelity and coverage; notably, starting from base samples with FID 3.95, latent space refinement achieves a \textbf{state-of-the-art} FID of \textbf{1.46} on MNIST using only a single additional function evaluation (1-NFE), while maintaining sample diversity.
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。