FADE: Frame-Aware Diffusion-Transformer-based Multi-Concept Erasure for Video Unlearning

2026-10-02

AI Summary

This study addresses the challenges of frame-level residual leakage and concurrent multi-concept interference in concept erasure for video generation models by proposing the FADE framework. To the best of our knowledge, this work is the first to introduce a gating mechanism over frame indices and denoising timesteps. By integrating closed-form editing, frame-aware low-rank adapters, and hard negative training, FADE achieves precise disentanglement and elimination of multiple concepts. Experimental results demonstrate that FADE reduces residual accuracy to 4.9% across 16 tasks, significantly outperforming existing baselines while maintaining stable VBench scores. Furthermore, the proposed approach generalizes effectively across various mainstream Diffusion Transformer architectures.

Abstract

Text-to-video (T2V) diffusion models can reproduce copyrighted, violent, or explicit content, which motivates concept erasure: removing designated concepts from a pretrained model while preserving its behavior on everything else. Existing T2V erasure methods leave two problems open. Their frame-agnostic suppression can leave isolated frames in which an erased concept resurfaces, a frame-reactivation gap that clip-level averages obscure; and they are usually evaluated with one target concept or category at a time. We propose Frame-Aware Diffusion Erasure (FADE), a multi-concept video unlearning framework. FADE first applies a joint closed-form key/value edit that suppresses all target concepts, then trains per-concept frame-aware low-rank adapters whose strength is gated by the frame index and the denoising timestep to remove residual per-frame leakage. Each adapter is trained with the other targets'prompts as hard negatives, which keeps the concept-specific components of different adapters well separated, and a similarity-based soft router combines the adapters according to the prompt. With 16 concepts (objects, artistic styles, and nudity) erased from a single Wan2.1-T2V-1.3B backbone, FADE reduces the residual accuracy on the object benchmark to 4.9%, against 15.5% for the strongest of eight baselines, while keeping the VBench average within 0.9% of the unedited model. The ranking is unchanged under a VLM judge and a blinded human study, and the advantage over the strongest baseline carries over to prompts that combine several erased concepts, to 30 simultaneously erased celebrity identities, and to Wan2.1-T2V-14B, CogVideoX-2B, and HunyuanVideo-1.5.

Problem

• concept erasure
• video unlearning
• text-to-video diffusion models
• frame-reactivation
• multi-concept erasure

Innovation

• Video Unlearning
• Multi-Concept Erasure
• Frame-Aware LoRA
• Diffusion Transformer
• Soft Router