ConceptPrism：通过残差词元优化实现个性化扩散模型中的概念解耦
ConceptPrism: Concept Disentanglement in Personalized Diffusion Models via Residual Token Optimization
通过跨图像比较与残差词元优化实现概念解耦，可直接用于个性化 T2I 中抑制参考图像里与用户提示冲突的无关属性，提升概念保真度与文本对齐的平衡。
原文依据 We jointly optimize a target token and image-wise residual tokens via reconstruction and exclusion losses.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论个性化文本到图像（T2I）生成已成为一种关键应用，能够从少量参考图像中创建用户特定的概念。其核心挑战是概念解耦：将目标概念与无关的残差信息分离。若缺乏这种解耦，在捕获高保真特征时往往会引入与用户提示相冲突的 undesired 属性，从而损害概念保真度与文本对齐之间的权衡。现有方法依赖人工引导，往往难以表示复杂的视觉细节，且缺乏可扩展性。我们提出 ConceptPrism，该框架仅通过跨图像比较提取共享特征，无需外部信息。我们通过重建损失与排除损失联合优化一个目标词元与逐图像残差词元。排除损失通过抑制残差词元中的共享信息，制造出一个信息真空，迫使目标词元捕获共同概念。大量评估表明，ConceptPrism 实现了准确的概念解耦，并在多样且复杂的视觉概念上显著提升了整体性能。
英文原文摘要 1185 字符
Personalized text-to-image (T2I) generation has emerged as a key application for creating user-specific concepts from a few reference images. The core challenge is concept disentanglement: separating the target concept from irrelevant residual information. Lacking such disentanglement, capturing high-fidelity features often incorporates undesired attributes that conflict with user prompts, compromising the trade-off between concept fidelity and text alignment. While existing methods rely on manual guidance, they often fail to represent intricate visual details and lack scalability. We introduce ConceptPrism, a framework that extracts shared features exclusively through cross-image comparison without external information. We jointly optimize a target token and image-wise residual tokens via reconstruction and exclusion losses. By suppressing shared information in residual tokens, the exclusion loss creates an information vacuum that forces the target token to capture the common concept. Extensive evaluations demonstrate that ConceptPrism achieves accurate concept disentanglement and significantly improves overall performance across diverse and complex visual concepts.