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

Text-to-video (T2V) diffusion models can generate realistic depictions of actions such as kicking, stabbing, and shooting, raising safety concerns that motivate targeted concept erasure. Although concept erasure has been extensively studied for static concepts in text-to-image and T2V models, erasing motion concepts remains largely unexplored. We present a systematic study of motion concept erasure in video Diffusion Transformers (DiTs). Through causal interventions, we show that text-conditioning attention carries concept-specific motion information and supports selective intervention, whereas perturbing temporal positional encoding suppresses both target and non-target dynamics. We further find that directly adapting ESD, a representative weight-level image erasure method, to a video DiT yields modest and uneven motion suppression: reducing its erasure training loss does not by itself remove the concept signal from the difference between the conditional and unconditional predictions, which classifier-free guidance (CFG) then scales at every denoising step. From these findings, we derive three requirements for motion concept erasure: concept specificity, spatial selectivity, and temporal naturalness. Each determines one component of MUTE (Motion concept Unlearning in Text-to-video gEneration): at each denoising step, MUTE extracts a concept direction through token neutralization, derives a spatial gate from the direction's intrinsic structure, and subtracts the resulting correction from the velocity output before CFG is applied. MUTE is training-free and requires no weight modification. Experiments on 20 motion concepts show that MUTE outperforms representative prompt-level, weight-level, and inference-time baselines on Wan2.1-T2V, and the same formulation transfers to CogVideoX, supporting its applicability across distinct T2V attention architectures.

Concept erasure techniques for text-to-video (T2V) diffusion models report substantial suppression of sensitive content, yet current evaluation is limited to checking whether the target concept is absent from generated frames, treating output-level suppression as evidence of representational removal. We introduce PROBE, a diagnostic protocol that quantifies the reactivation potential of erased concepts in T2V models. With all model parameters frozen, PROBE optimizes a lightweight pseudo-token embedding through a denoising reconstruction objective combined with a novel latent alignment constraint that anchors recovery to the spatiotemporal structure of the original concept. We make three contributions: (1) a multi-level evaluation framework spanning classifier-based detection, semantic similarity, temporal reactivation analysis, and human validation; (2) systematic experiments across three T2V architectures, three concept categories, and three erasure strategies revealing that all tested methods leave measurable residual capacity whose robustness correlates with intervention depth; and (3) the identification of temporal re-emergence, a video-specific failure mode where suppressed concepts progressively resurface across frames, invisible to frame-level metrics. These findings suggest that current erasure methods achieve output-level suppression rather than representational removal. We release our protocol to support reproducible safety auditing. Our code is available at https://github.com/YiweiXie/PRObingBasedEvaluation.

Removing undesired concepts from large-scale text-to-image (T2I) and text-to-video (T2V) diffusion models while preserving overall generative quality remains a major challenge, particularly as modern models such as Stable Diffusion v3, Flux, and OpenSora employ flow-matching and transformer-based architectures and extend to long-horizon video generation. Existing concept erasure methods, designed for earlier T2I/T2V models, often fail to generalize to these paradigms. To address this issue, we propose EraseAnything++, a unified framework for concept erasure in both image and video diffusion models with flow-matching objectives. Central to our approach is formulating concept erasure as a constrained multi-objective optimization problem that explicitly balances concept removal with preservation of generative utility. To solve the resulting conflicting objectives, we introduce an efficient utility-preserving unlearning strategy based on implicit gradient surgery. Furthermore, by integrating LoRA-based parameter tuning with attention-level regularization, our method anchors erasure on key visual representations and propagates it consistently across spatial and temporal dimensions. In the video setting, we further enhance consistency through an anchor-and-propagate mechanism that initializes erasure on reference frames and enforces it throughout subsequent transformer layers, thereby mitigating temporal drift. Extensive experiments on both image and video benchmarks demonstrate that EraseAnything++ substantially outperforms prior methods in erasure effectiveness, generative fidelity, and temporal consistency, establishing a new state of the art for concept erasure in next-generation diffusion models.