arXiv:2510.14376v2 Announce Type: replace 
Abstract: Recent progress in text-to-image (T2I) generative models has led to significant improvements in generating high-quality images aligned with text prompts. However, these models still struggle with prompts involving multiple objects, often resulting in object neglect or object mixing. Through extensive studies, we identify four problematic scenarios, Similar Shapes, Similar Textures, Dissimilar Background Biases, and Many Objects, where inter-object relationships frequently lead to such failures. Motivated by two key observations about CLIP embeddings, we propose DOS (Directional Object Separation), a method that modifies three types of CLIP text embeddings before passing them into text-to-image models. Experimental results show that DOS consistently improves the success rate of multi-object image generation and reduces object mixing. In human evaluations, DOS significantly outperforms four competing methods, receiving 26.24%-43.04% more votes across four benchmarks. These results highlight DOS as a practical and effective solution for improving multi-object image generation.

تسلط دراسة حديثة حول نماذج توليد الصور من النصوص الضوء على التقدم في إنشاء صور عالية الجودة من العبارات النصية. ومع ذلك، لا تزال هناك تحديات عندما تتضمن العبارات عدة أشياء، مما يؤدي إلى مشكلات مثل إهمال الأشياء أو خلطها. تحدد الدراسة أربعة سيناريوهات رئيسية تعاني من مشاكل تؤثر على توليد الصور، مما يمهد الطريق لتقنيات محسّنة في عرض الأشياء المتعددة. هذا مهم لأنه يعالج قيدًا شائعًا في الصور التي تنتجها الذكاء الاصطناعي، مما يعزز الإمكانية لإنشاء محتوى بصري أكثر دقة وتفصيلاً.

Un estudio reciente sobre modelos generativos de texto a imagen destaca los avances en la creación de imágenes de alta calidad a partir de indicaciones textuales. Sin embargo, persisten desafíos cuando las indicaciones involucran múltiples objetos, lo que lleva a problemas como la negligencia o la mezcla de objetos. La investigación identifica cuatro escenarios problemáticos clave que afectan la generación de imágenes, allanando el camino para técnicas mejoradas en el renderizado de múltiples objetos. Esto es significativo ya que aborda una limitación común en la imagen generada por IA, mejorando el potencial para contenido visual más preciso y detallado.

Une étude récente sur les modèles génératifs de texte à image met en lumière les avancées dans la création d'images de haute qualité à partir de descriptions textuelles. Cependant, des défis subsistent lorsque les descriptions impliquent plusieurs objets, entraînant des problèmes tels que la négligence ou le mélange d'objets. La recherche identifie quatre scénarios problématiques clés qui affectent la génération d'images, ouvrant la voie à des techniques améliorées pour le rendu multi-objets. Cela est significatif car cela aborde une limitation courante dans l'imagerie générée par l'IA, améliorant le potentiel pour un contenu visuel plus précis et détaillé.

A recent study on text-to-image generative models highlights advancements in creating high-quality images from text prompts. However, challenges remain when prompts involve multiple objects, leading to issues like object neglect or mixing. The research identifies four key problematic scenarios that affect image generation, paving the way for improved techniques in multi-object rendering. This is significant as it addresses a common limitation in AI-generated imagery, enhancing the potential for more accurate and detailed visual content.

DOS: Directional Object Separation in Text Embeddings for Multi-Object Image Generation

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.

DOS: Directional Object Separation in Text Embeddings for Multi-Object Image Generation

Was this article worth reading? Share it

4o Image Gen

WasItAI

DrawMy.Pet