“I think the paper is well worth the read, it's not particularly long (much is references and appendices), and nicely written”
Extracting Training Data from Diffusion Models
扩散模型会记住并复现训练图像,且可被提取
论文证明扩散模型会记住训练集中的个别图像,并在生成时原样输出。作者用「生成再过滤」的流程,从最先进的模型中提取出一千多个训练样本,包括真人照片和注册商标 logo;还训练了数百个不同设置的扩散模型,分析建模与数据选择如何影响隐私。总体结论是扩散模型的隐私性远不如 GAN 等早期生成模型,缓解这些漏洞可能需要隐私保护训练方面的新进展。
评论者在争论 AI 是否会记忆、复制训练数据以及版权问题时,把它当作扩散模型会复现训练图像的实证依据,也有人质疑其用刻意构造的提示来提取。
HN 书评
推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及“For those following along, (1) is the best paper I could find talking about extracting images from SD.”
“for papers like this, where they're generating hundreds of millions of images and digging into CLIP embeddings to construct prompts that intentionally recreate images”
“In this work, we show that diffusion models memorize individual images from their training data and emit them at generation time.”
“IANAL either, but there are attacks you can use to extract the original training data from the models: Extracting Training Data from Diffusion Models.”
“We already know that, at least for some image models, the answer to that question is”
看全部 15 条 →
“My mistake! I mis-remembered the original Google paper[0], which stated only that the image was in the Stable Diffusion training set.”
“often to the point where training data can be extracted [1][2]”
“There's a paper from Google and Princeton about regurgitation happening in Stable Diffusion and Imagen: Extracting Training Data from Diffusion Models”
“It's already proven that you can reconstruct at least a small fraction of the training set from diffusion models.”
“Look at the Ann Graham Lotz image and tell me that isn't the source image being reproduced in a lossy manner.”
“some exploits discovered lately seem to suggest that this isn't just "reading a page" and then moving on: Extracting Training Data from Diffusion Models”
“Here's one-- Extracting Training Data from Diffusion Models Extracting Training Data from Diffusion Models”
“Producing memorized images has precedent[0].”
“But Ai model have been shown to reproduce the training data”
- Extracting Training Data from Diffusion Models ↗tbruckner · 2023-01-31 · 3 分
- Stable Diffusion Memorizes Training Points ↗ericwallace_ucb · 2023-01-31 · 1 分