arXiv:2510.22851v1 Announce Type: new 
Abstract: Concept erasure in text-to-image diffusion models is crucial for mitigating harmful content, yet existing methods often compromise generative quality. We introduce Semantic Surgery, a novel training-free, zero-shot framework for concept erasure that operates directly on text embeddings before the diffusion process. It dynamically estimates the presence of target concepts in a prompt and performs a calibrated vector subtraction to neutralize their influence at the source, enhancing both erasure completeness and locality. The framework includes a Co-Occurrence Encoding module for robust multi-concept erasure and a visual feedback loop to address latent concept persistence. As a training-free method, Semantic Surgery adapts dynamically to each prompt, ensuring precise interventions. Extensive experiments on object, explicit content, artistic style, and multi-celebrity erasure tasks show our method significantly outperforms state-of-the-art approaches. We achieve superior completeness and robustness while preserving locality and image quality (e.g., 93.58 H-score in object erasure, reducing explicit content to just 1 instance, and 8.09 H_a in style erasure with no quality degradation). This robustness also allows our framework to function as a built-in threat detection system, offering a practical solution for safer text-to-image generation.

تم تقديم إطار جديد يسمى Semantic Surgery لتحسين إزالة المفاهيم في نماذج الانتشار من النص إلى الصورة دون المساس بجودة التوليد. تعمل هذه الطريقة المبتكرة مباشرة على تضمينات النص، مما يسمح بطريقة أكثر فعالية لتحييد المحتوى الضار في الصور المولدة. هذا الأمر مهم لأنه يعالج تحديًا كبيرًا في المحتوى الذي يتم إنشاؤه بواسطة الذكاء الاصطناعي، مما يضمن استخدامًا أكثر أمانًا ومسؤولية للتكنولوجيا.

Se ha presentado un nuevo marco llamado Semantic Surgery para mejorar la eliminación de conceptos en modelos de difusión de texto a imagen sin comprometer la calidad generativa. Este enfoque innovador opera directamente sobre las incrustaciones de texto, permitiendo una forma más efectiva de neutralizar contenido dañino en las imágenes generadas. Esto es importante porque aborda un desafío significativo en el contenido generado por IA, asegurando un uso más seguro y responsable de la tecnología.

Un nouveau cadre appelé Semantic Surgery a été introduit pour améliorer l'effacement des concepts dans les modèles de diffusion texte-image sans compromettre la qualité générative. Cette approche innovante agit directement sur les embeddings de texte, permettant une manière plus efficace de neutraliser le contenu nuisible dans les images générées. Cela est important car cela répond à un défi majeur dans le contenu généré par l'IA, garantissant une utilisation plus sûre et responsable de la technologie.

A new framework called Semantic Surgery has been introduced to improve concept erasure in text-to-image diffusion models without compromising generative quality. This innovative approach operates directly on text embeddings, allowing for a more effective way to neutralize harmful content in generated images. This matters because it addresses a significant challenge in AI-generated content, ensuring safer and more responsible use of technology.

Semantic Surgery: Zero-Shot Concept Erasure in Diffusion Models

Was this article worth reading? Share it

Humanize AI

LucidQuery AI

Semantic Pen

Blunge

SVGenius

4o Image Gen

Ready to build your own newsroom?