发现扩散模型高密度区生成卡通和模糊图像，提出高效采样新方法。
Diffusion Models as Cartoonists: The Curious Case of High Density Regions
- 通过模式追踪技术精准定位去噪分布的高似然模式。
- 新采样器生成的图像似然值显著高于传统方法，且多为卡通或模糊图。
- 无需额外计算成本即可追踪采样似然，适合对生成质量敏感的研究者。
我们研究了扩散模型高密度区域所包含的图像类型。提出一种理论上的模式追踪过程，可精确定位去噪分布的模式，并设计了一种实用的高密度采样方法，其生成的图像似然值持续高于常规采样器。实验发现，典型采样器无法生成的高似然样本真实存在，常表现为卡通画或模糊图像，且这些模式出现在不含此类样本的数据集中。此外，我们提出一种新颖的扩散SDE采样似然追踪方法，计算开销几乎为零。代码已公开于 https://github.com/Aalto-QuML/high-density-diffusion。
原文摘要 · Abstract (English)
We investigate what kind of images lie in the high-density regions of diffusion models. We introduce a theoretical mode-tracking process capable of pinpointing the exact mode of the denoising distribution, and we propose a practical high-density sampler that consistently generates images of higher likelihood than usual samplers. Our empirical findings reveal the existence of significantly higher likelihood samples that typical samplers do not produce, often manifesting as cartoon-like drawings or blurry images depending on the noise level. Curiously, these patterns emerge in datasets devoid of such examples. We also present a novel approach to track sample likelihoods in diffusion SDEs, which remarkably incurs no additional computational cost. Code is available at https://github.com/Aalto-QuML/high-density-diffusion.
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。