arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2407.21720cs.CV
收藏

扩散模型中记忆的检测、解释与缓解

Detecting, Explaining, and Mitigating Memorization in Diffusion Models

University of Maryland(马里兰大学);Sony AI(索尼人工智能公司);Zhejiang University(浙江大学)

Yuxin Wen, Yuchen Liu, Chen Chen, Lingjuan Lyu

AI总结:

本文提出通过文本条件预测幅度检测扩散模型记忆现象,并进一步给出可解释分析与推理最小化、训练过滤两种缓解策略,在保持生成质量的同时降低训练数据复制风险。

摘要:

扩散模型的最新突破展现出卓越的图像生成能力。然而,研究表明,一些输出仅仅是对训练数据的复制。这种复制给模型所有者带来了潜在的法律挑战,尤其是当生成内容包含专有信息时。在这项工作中,我们引入了一种简单而有效的方法,通过检查文本条件预测的幅度来检测被记忆的提示。我们提出的方法能够无缝集成而不干扰采样算法,并且即使在第一个生成步骤、每个提示仅生成一次的情况下也能提供高精度。基于我们的检测策略,我们揭示了一种可解释的方法,展示单个单词或标记对记忆的贡献。这为用户调整其提示提供了一种交互式媒介。此外,我们提出了两种策略,即通过利用文本条件预测的幅度来缓解记忆,具体方式为在推理过程中进行最小化或在训练过程中进行过滤。这些提出的策略在保持高生成质量的同时有效抵消了记忆现象。代码可在 https://github.com/YuxinWenRick/diffusion_memorization 获取。

英文摘要:

Recent breakthroughs in diffusion models have exhibited exceptional image-generation capabilities. However, studies show that some outputs are merely replications of training data. Such replications present potential legal challenges for model owners, especially when the generated content contains proprietary information. In this work, we introduce a straightforward yet effective method for detecting memorized prompts by inspecting the magnitude of text-conditional predictions. Our proposed method seamlessly integrates without disrupting sampling algorithms, and delivers high accuracy even at the first generation step, with a single generation per prompt. Building on our detection strategy, we unveil an explainable approach that shows the contribution of individual words or tokens to memorization. This offers an interactive medium for users to adjust their prompts. Moreover, we propose two strategies i.e., to mitigate memorization by leveraging the magnitude of text-conditional predictions, either through minimization during inference or filtering during training. These proposed strategies effectively counteract memorization while maintaining high-generation quality. Code is available at https://github.com/YuxinWenRick/diffusion_memorization.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑