DICE：离散扩散与掩码生成模型的离散逆向方法
计算机视觉与模式识别
2025-11-14 v3 机器学习
摘要
离散扩散模型在图像生成和掩码语言建模等任务中取得了成功，但在受控内容编辑方面面临局限。我们提出DICE（Discrete Inversion for Controllable Editing），这是首个实现离散扩散模型精确逆向方法的方案，包括多项式扩散和掩码生成模型。通过记录逆扩散过程中的噪声序列和掩码模式，DICE enables accurate reconstruction and flexible editing of discrete data without the need for predefined masks or attention manipulation. 我们在图像和文本域 demonstrate了 DICE的有效性，在 VQ-Diffusion、Paella 和 RoBERTa 等模型上进行评估。我们的结果表明，DICE 在保持高数据保真度的同时提升了编辑能力，为离散空间中的细粒度内容操控提供了新机会。
引用
@article{arxiv.2410.08207,
title = {DICE: Discrete Inversion Enabling Controllable Editing for Multinomial Diffusion and Masked Generative Models},
author = {Xiaoxiao He and Quan Dao and Ligong Han and Song Wen and Minhao Bai and Di Liu and Han Zhang and Martin Renqiang Min and Felix Juefei-Xu and Chaowei Tan and Bo Liu and Kang Li and Hongdong Li and Junzhou Huang and Faez Ahmed and Akash Srivastava and Dimitris Metaxas},
journal= {arXiv preprint arXiv:2410.08207},
year = {2025}
}
备注
Project webpage: https://hexiaoxiao-cs.github.io/DICE/. This paper was accepted to CVPR 2025 but later desk-rejected post camera-ready, due to a withdrawal from ICLR made 14 days before reviewer assignment