发现扩散模型采样器会放大或减弱数据偏见，关键在超参数设计。
How I Met Your Bias: Investigating Bias Amplification in Diffusion Models
- 通过控制实验揭示采样算法影响偏见传播程度。
- 同一训练模型下，不同采样设置可使偏见增强或减弱。
- 对生成式AI伦理评估有指导意义，适合关注公平性的研究者。
基于扩散的生成模型在图像合成任务中表现卓越，但其复制并放大数据集偏见的机制尚不明确。尽管以往研究认为偏见放大是扩散模型固有特性，本文首次系统分析采样算法及其超参数对偏见放大效应的影响。通过在Biased MNIST、Multi-Color MNIST和BFFHQ数据集上，以及Stable Diffusion模型上的受控实验，我们证实采样器——通常为提升生成质量和速度而优化——对偏见放大具有显著且可测量的影响。即使训练模型固定，调整采样超参数仍可导致偏见的增强或抑制。源代码已公开于https://github.com/How-I-met-your-bias/how_i_met_your_bias。
原文摘要 · Abstract (English)
Diffusion-based generative models demonstrate state-of-the-art performance across various image synthesis tasks, yet their tendency to replicate and amplify dataset biases remains poorly understood. Although previous research has viewed bias amplification as an inherent characteristic of diffusion models, this work provides the first analysis of how sampling algorithms and their hyperparameters influence bias amplification. We empirically demonstrate that samplers for diffusion models -- commonly optimized for sample quality and speed -- have a significant and measurable effect on bias amplification. Through controlled studies with models trained on Biased MNIST, Multi-Color MNIST and BFFHQ, and with Stable Diffusion, we show that sampling hyperparameters can induce both bias reduction and amplification, even when the trained model is fixed. Source code is available at https://github.com/How-I-met-your-bias/how_i_met_your_bias.
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。