arXiv:2606.24192v1 Announce Type: new 
Abstract: Unlearning has emerged as a key technique to mitigate harmful content generation in diffusion models. However, existing methods often remove not only the target concept, but also benign co-occurring concepts. As illustrated in Fig.1, unlearning nudity can unintentionally suppress the concept of person, preventing a model from generating images with person. We define these undesirably suppressed co-occurring concepts that must be preserved CARE (Co-occurring Associated REtained concepts). Then, we introduce the CARE score, a general metric that directly quantifies their preservation across unlearning tasks. With this foundation, we propose ReCARE (Robust erasure for CARE), a framework that explicitly safeguards CARE while erasing only the target concept. ReCARE automatically constructs the CARE-set, a curated vocabulary of benign co-occurring tokens extracted from target images, and leverages this vocabulary during training for stable unlearning. Extensive experiments across various target concepts (Nudity, Van Gogh style, and Tench object) demonstrate that ReCARE achieves overall state-of-the-art performance in balancing robust concept erasure, overall utility, and CARE preservation.

دراسة حديثة حول تقنيات نسيان المحتوى في نماذج الانتشار تسلط الضوء على التحدي المتمثل في الحفاظ على المفاهيم المتزامنة الحميدة أثناء إزالة المحتوى الضار. تقدم الدراسة مقياس CARE وإطار ReCARE، الذي يهدف إلى حماية هذه المفاهيم خلال عملية النسيان.

Un estudio reciente sobre técnicas de desaprendizaje en modelos de difusión destaca el desafío de preservar conceptos co-ocurrentes benignos mientras se elimina contenido dañino. La investigación introduce el puntaje CARE y el marco ReCARE, que busca salvaguardar estos conceptos durante el proceso de desaprendizaje.

Une étude récente sur les techniques d'apprentissage inversé dans les modèles de diffusion met en lumière le défi de préserver les concepts co-occurrents bénins tout en éliminant le contenu nuisible. La recherche introduit le score CARE et le cadre ReCARE, qui vise à protéger ces concepts durant le processus d'apprentissage inversé.

A recent study on unlearning techniques in diffusion models highlights the challenge of preserving benign co-occurring concepts while removing harmful content. The research introduces the CARE score and the ReCARE framework, which aims to safeguard these concepts during the unlearning process.