평범한 필기장

Paper : https://arxiv.org/abs/2412.20413 EraseAnything: Enabling Concept Erasure in Rectified Flow TransformersRemoving unwanted concepts from large-scale text-to-image (T2I) diffusion models while maintaining their overall generative quality remains an open challenge. This difficulty is especially pronounced in emerging paradigms, such as Stable Diffusion (SD) v3arxiv.org0. Obstacles in migrati..

Paper : https://arxiv.org/abs/2412.09611 FluxSpace: Disentangled Semantic Editing in Rectified Flow TransformersRectified flow models have emerged as a dominant approach in image generation, showcasing impressive capabilities in high-quality image synthesis. However, despite their effectiveness in visual generation, rectified flow models often struggle with disentanarxiv.orgAbstract기존 문제 Rectifi..