EraseAnything++: Enabling Concept Erasure in Rectified Flow Transformers Leveraging Multi-Object Optimization

Researchers introduced EraseAnything++, a new framework for removing unwanted concepts from advanced AI image and video generation models like Stable Diffusion v3 and Flux. The method uses multi-objective optimization to balance concept removal while preserving overall generative quality, showing superior performance compared to existing approaches.

• →EraseAnything++ addresses concept erasure limitations in modern flow-matching and transformer-based diffusion models.
• →The framework formulates concept erasure as a constrained multi-objective optimization problem balancing removal with quality preservation.
• →The method introduces utility-preserving unlearning strategy based on implicit gradient surgery for conflicting objectives.
• →Integration of LoRA-based parameter tuning with attention-level regularization anchors erasure on key visual representations.
• →Extensive experiments demonstrate substantial outperformance in erasure effectiveness, generative fidelity, and temporal consistency.