arXiv:2510.24820v1 Announce Type: new 
Abstract: With the rapid advancement of text-to-image (T2I) models, ensuring their safety has become increasingly critical. Existing safety approaches can be categorized into training-time and inference-time methods. While inference-time methods are widely adopted due to their cost-effectiveness, they often suffer from limitations such as over-refusal and imbalance between safety and utility. To address these challenges, we propose a multi-round safety editing framework that functions as a model-agnostic, plug-and-play module, enabling efficient safety alignment for any text-to-image model. Central to this framework is MR-SafeEdit, a multi-round image-text interleaved dataset specifically constructed for safety editing in text-to-image generation. We introduce a post-hoc safety editing paradigm that mirrors the human cognitive process of identifying and refining unsafe content. To instantiate this paradigm, we develop SafeEditor, a unified MLLM capable of multi-round safety editing on generated images. Experimental results show that SafeEditor surpasses prior safety approaches by reducing over-refusal while achieving a more favorable safety-utility balance.

تقديم SafeEditor يمثل تقدمًا كبيرًا في مجال تحرير أمان نماذج النص إلى صورة (T2I). مع تزايد انتشار نماذج T2I، أصبح ضمان أمانها أمرًا حيويًا، ويقدم SafeEditor نهجًا موحدًا لمعالجة القيود الموجودة في الأساليب الحالية. من خلال التركيز على تعديلات الأمان متعددة الجولات، يهدف إلى تحسين كل من الأمان والفائدة، مما يجعله أداة قيمة للمطورين والمستخدمين على حد سواء. لا تعالج هذه الابتكار المخاوف الأمنية الملحة فحسب، بل تمهد أيضًا الطريق لتطبيقات T2I أكثر موثوقية وفعالية.

La introducción de SafeEditor marca un avance significativo en el campo de la edición de seguridad para modelos de texto a imagen (T2I). A medida que los modelos T2I se vuelven más comunes, garantizar su seguridad es crucial, y SafeEditor ofrece un enfoque unificado para abordar las limitaciones de los métodos actuales. Al centrarse en ediciones de seguridad en múltiples rondas, busca mejorar tanto la seguridad como la utilidad, convirtiéndose en una herramienta valiosa para desarrolladores y usuarios. Esta innovación no solo aborda preocupaciones de seguridad urgentes, sino que también allana el camino para aplicaciones T2I más confiables y efectivas.

L'introduction de SafeEditor représente une avancée significative dans le domaine de l'édition de sécurité pour les modèles de texte à image (T2I). Alors que les modèles T2I deviennent de plus en plus courants, garantir leur sécurité est crucial, et SafeEditor propose une approche unifiée pour s'attaquer aux limitations des méthodes actuelles. En se concentrant sur des modifications de sécurité en plusieurs étapes, il vise à améliorer à la fois la sécurité et l'utilité, ce qui en fait un outil précieux pour les développeurs et les utilisateurs. Cette innovation répond non seulement aux préoccupations de sécurité pressantes, mais ouvre également la voie à des applications T2I plus fiables et efficaces.

The introduction of SafeEditor marks a significant advancement in the field of text-to-image (T2I) safety editing. As T2I models become more prevalent, ensuring their safety is crucial, and SafeEditor offers a unified approach to tackle existing limitations in current methods. By focusing on multi-round safety edits, it aims to enhance both safety and utility, making it a valuable tool for developers and users alike. This innovation not only addresses pressing safety concerns but also paves the way for more reliable and effective T2I applications.

SafeEditor: Unified MLLM for Efficient Post-hoc T2I Safety Editing

arXiv:2511.18684v1 Announce Type: new 
Abstract: Robust concept removal for text-to-image (T2I) and text-to-video (T2V) models is essential for their safe deployment. Existing methods, however, suffer from costly retraining, inference overhead, or vulnerability to adversarial attacks. Crucially, they rarely model the latent semantic overlap between the target erase concept and surrounding content -- causing collateral damage post-erasure -- and even fewer methods work reliably across both T2I and T2V domains. We introduce Instant Concept Erasure (ICE), a training-free, modality-agnostic, one-shot weight modification approach that achieves precise, persistent unlearning with zero overhead. ICE defines erase and preserve subspaces using anisotropic energy-weighted scaling, then explicitly regularises against their intersection using a unique, closed-form overlap projector. We pose a convex and Lipschitz-bounded Spectral Unlearning Objective, balancing erasure fidelity and intersection preservation, that admits a stable and unique analytical solution. This solution defines a dissociation operator that is translated to the model's text-conditioning layers, making the edit permanent and runtime-free. Across targeted removals of artistic styles, objects, identities, and explicit content, ICE efficiently achieves strong erasure with improved robustness to red-teaming, all while causing only minimal degradation of original generative abilities in both T2I and T2V models.

قدم الباحثون تقنية Instant Concept Erasure (ICE)، وهي نهج جديد لإزالة المفاهيم بشكل موثوق في نماذج النص إلى صورة (T2I) والنص إلى فيديو (T2V). تلغي هذه الطريقة الحاجة إلى إعادة التدريب المكلفة وتقلل من عبء الاستدلال، بينما تعالج الثغرات الأمنية ضد الهجمات العدائية. تستخدم ICE تقنية تعديل الوزن في خطوة واحدة دون تدريب، مما يضمن إلغاء تعلم دقيق ودائم دون أضرار جانبية للمحتوى المحيط.

Investigadores han presentado el Instant Concept Erasure (ICE), un enfoque novedoso para la eliminación robusta de conceptos en modelos de texto-a-imagen (T2I) y texto-a-video (T2V). Este método elimina la necesidad de reentrenamiento costoso y minimiza la sobrecarga de inferencia, al tiempo que aborda las vulnerabilidades a ataques adversariales. ICE emplea una técnica de modificación de pesos en una sola vez, sin necesidad de entrenamiento, que asegura un desaprendizaje preciso y persistente sin daños colaterales al contenido circundante.

Des chercheurs ont introduit l'Instant Concept Erasure (ICE), une nouvelle approche pour la suppression robuste de concepts dans les modèles de texte-à-image (T2I) et de texte-à-vidéo (T2V). Cette méthode élimine le besoin de réentraînement coûteux et minimise la surcharge d'inférence tout en abordant les vulnérabilités aux attaques adversariales. ICE utilise une technique de modification de poids en une seule fois, sans formation, qui garantit un désapprentissage précis et persistant sans dommages collatéraux au contenu environnant.

Researchers have introduced Instant Concept Erasure (ICE), a novel approach for robust concept removal in text-to-image (T2I) and text-to-video (T2V) models. This method eliminates the need for costly retraining and minimizes inference overhead while addressing vulnerabilities to adversarial attacks. ICE employs a training-free, one-shot weight modification technique that ensures precise and persistent unlearning without collateral damage to surrounding content.

SafeEditor: Unified MLLM for Efficient Post-hoc T2I Safety Editing

Was this article worth reading? Share it

URLtoText

ComfyUI

Humanize AI