Overview

Concept erasure studies how to suppress targeted information—such as unsafe content, personal identities, or artistic styles—without damaging a model’s general capabilities. In generative models, common approaches fine-tune weights or use inference-time steering, while representation-based methods identify and modify concept directions, subspaces, or attention pathways. Researchers test these methods on image and video generation, language models, and other settings, measuring both target leakage and preservation of unrelated outputs, including under adversarial prompts. A central challenge is that erased concepts can remain recoverable through alternate representations or prompts, while stronger suppression can cause collateral changes, so reliable evaluation and balancing removal against utility remain open problems.

Latest papers

70 papers8 papers in the last four weeks, up 33% on the four weeks before. 0.1% of all new papers.

Nov 16, 2026

Example-Based Spatial Guidance for Training-Free Concept Erasure in Diffusion Models

Text-to-image (T2I) diffusion models often synthesize policy-violating content, necessitating robust safeguards and rigorous evaluation frameworks. However, current approaches remain largely monolithic in both execution and evaluation, limited to coarse-grained categorization within broad unsafe concepts. Such approaches fail to account for the specific constituent factors of a violation, resulting in weak learning signals and allowing model collapse to bypass safety checks. To overcome these limitations, we propose Example-Based Spatial Guidance (EBSG), a training-free method that utilizes user-editable exemplar packs to provide granular spatial guidance for precise concept steering. Unlike existing methods that coarsely steer the entire image away from a broad concept, EBSG decomposes these categories into specific sub-concepts via exemplar text-image pairs, providing explicit, localized steering signals for more robust safety control. Furthermore, we introduce a vision language model (VLM)-based evaluation protocol that provides a fine-grained assessment, avoiding the pitfall of conventional binary evaluators that permit model collapse to bypass safety checks. Empirically, EBSG achieves state-of-the-art erasure performance across diverse safety datasets and concept categories, spanning four nudity benchmarks, seven I2P harmful-concept slices, and four MJA metaphor categories, while supporting multi-concept removal and transferring to a modern backbone, SD3.
Nov 11, 2026

From Concepts to Interactions: Dynamic Graph Semantic Recomposition for Safe Visual Generation

Recent advances in flow-matching Transformers have greatly improved Text-to-Image (T2I) and Text-to-Video (T2V) generation, but unconstrained generation risks producing illegal or unethical content. Safety alignment must therefore suppress unsafe content while preserving legitimate generation. A concept's safety depends on context: a knife used for cooking is safe, whereas its use in violence is harmful. Yet most concept erasure methods indiscriminately suppress static, isolated representations, overlooking evolving semantic interactions. Consequently, safe scenes containing potentially sensitive terms may be unintentionally suppressed, compromising generation quality. To address this, we propose Dynamic Graph Semantic Recomposition (DGSR), which dynamically constructs and updates a semantic interaction graph from evolving text visual states during sampling. Context aware graph reasoning captures semantic dependencies, while paired safe/unsafe supervision distinguishes harmful from legitimate interactions, enabling precise identification of and selective intervention on high risk semantic relationships while preserving benign semantics. DGSR freezes the pretrained backbone and supports both image and video generation. Experiments show that DGSR achieves the lowest Attack Success Rate (ASR) in 10 of 12 T2I evaluations and reduces the average T2V unsafe frame rate by 41.2% relative to the strongest baseline, while maintaining high generation fidelity and temporal consistency.
Nov 8, 2026

What Travels Across Languages? Reading, Transfer and Erasure of Cross-Lingual Concept Subspaces in Bangla Idioms

Cross-lingual concept analysis requires distinguishing the information that a subspace retains from the classifiers and erasers that transfer through it. We study these operations in six frozen encoders using 10,359 Bangla (Bengali) idiom senses, eight input conditions, and figurative-class and sentiment labels. Class-mean directions provide the most consistent evidence of shared coarse meaning: in post hoc analyses, they outperform random-subspace, permuted-label and covariance-matched nulls in all 12 model-label tests under balanced accuracy. A prospective internal re-split supports idiom-alone sharing in 10 of 12 tests, spanning five encoders. The original confirmatory logistic-regression family rejects both planned nulls in 4 of 12 tests, where English directions gain only a few accuracy points above the majority-class floor (3.0-7.3). Classifier transfer depends on the input map and regularization: English-trained class-mean readouts beat both zero-shot nulls in 8 of 12 discovery tests, and stronger source-selected penalties improve transfer through logistic and LEACE directions. These readouts use Bangla-validation-selected layers and unlabeled target preprocessing. Input controls show greater coarse-class sharing for explicit glosses than for idiom counterparts. Reading, however, does not imply effective erasure: English-fitted LEACE removes less than one percentage point of Bangla figurative-class accuracy, compared with 4-11 points for Bangla-fitted LEACE; INLP shows a similar asymmetry. Pattern-filter analysis, covariance-weighted input maps and a binary Gaussian overlap bound explain how these operations can diverge. The analogous multiclass ordering is an empirical diagnostic, observed in 11 of 12 cells. The results support an estimator-sensitive account of cross-lingual concept sharing and motivate evaluating reading, classifier transfer, erasure and overlap separately.
Nov 4, 2026

Shaping & Monitoring Models through their Internal Representations

We propose Mechanistic Shaping, a unified framework for shaping models through their internal representations. Mechanistic Shaping is applied during training so that models follow specifications (rules) about which concepts they express, measured via their outputs or internal representations (e.g., no harmful content), and also which concepts they rely on, i.e., what they use to make their decisions (e.g., not a candidate's gender when screening a résumé). Across multiple domains (vision, medical imaging, language), scales (up to 32B parameters), and training paradigms (classification, SFT, DPO, GRPO), we find that Mechanistic Shaping substantially reduces targeted behaviors while preserving task performance. Since training against a representation can also modify it, we ask whether shaped concepts remain monitorable, a question that depends on the class of monitors considered. We therefore introduce a hierarchy of such classes, which separates a concept that escapes one probe from a concept that is truly gone. We characterize three geometric mechanisms of circumvention: the concept can drift away from the monitor, be erased from the layer, or persist in higher-order statistics as a nonlinear encoding, which we explain through a product-code construction. Although other mechanisms may arise in different settings or at larger scales, these three account for all of our observations, and each can be detected by a matching monitor. In practice, we find that shaped concepts largely remain monitorable: refitted linear probes recover the concept in most settings, and the original probe remains reliable whenever gradients do not flow through it. Taken together, these results point to the potential of using interpretability as a tool for training, and call for a correspondingly rigorous science of interpretability-based training.
Nov 3, 2026

Is Forgetting Unified? Exploring Concept Unlearning in Unified Multimodal Models

Unified multimodal models (UMMs) integrate image generation and visual understanding, but are existing single-pathway unlearning methods sufficient to suppress target concepts across both? In this work, we study cross-pathway concept unlearning. Our goal is to prevent a model from both generating images of a target concept and identifying it in input images, while preserving other capabilities. To this end, we introduce UUBench, a diagnostic benchmark covering 30 target concepts and 60 semantically related neighboring concepts. Experiments on representative UMMs show that the evaluated single-pathway methods achieve limited joint forgetting: generation-side editors strongly suppress target generation but yield limited forgetting in visual understanding, while understanding-side baselines struggle to achieve substantial forgetting even in their own pathway. Their direct composition also yields limited joint forgetting. We therefore propose UniForget, achieving joint forgetting rates of 80.0% on BAGEL-7B-MoT and 92.7% on Janus-Pro-7B, outperforming representative baselines. Further analysis reveals trade-offs between joint forgetting and capability preservation.
Oct 30, 2026

KINDER: Kernel-based Independence for Fair Representation Learning via Prototype-space Erasure

Concept erasure is a prominent approach to achieving fairness in machine learning through the removal of sensitive attributes from learned representations. Prior concept erasure methods typically define debiasing indirectly through the failure of a chosen adversary, probe, or fairness penalty, which makes the target of erasure dependent on a particular decoder family or optimization setup. We introduce KINDER, which directly removes sensitive-attribute information from intermediate representations with respect to a chosen family of prototype functionals, and is compatible with any deep model containing an intermediate feature space, independent of the architecture, task type, or training paradigm. KINDER operates in a random Fourier feature space, where it estimates a sensitive subspace from protected group prototypes and projects representations onto its orthogonal complement. Thus, it forms an explicit representation cleaning mechanism distinct from prior fairness control studies. We conduct extensive experiments across diverse settings and show the effectiveness of KINDER on (i) supervised and self-supervised settings, (ii) classification, regression, and image segmentation tasks, and (iii) diverse data modalities, including visual, textual, and tabular data. Code and trained models are available at: https://github.com/abtinmU/KINDER
Oct 23, 2026

Principled Nonlinear Concept Erasure and Steering

Learned representations often encode sensitive attributes, such as gender or race, that may be undesirable or unsafe to retain. Concept erasure aims to remove this information while preserving task-relevant content. Existing methods remain incomplete: linear methods provide guarantees only against linear probes, while nonlinear methods achieve stronger empirical erasure but lack formal guarantees and often sacrifice utility. In this work, we introduce a principled conditional autoencoder framework for nonlinear concept erasure. We follow the principle that erasure should remove only the sensitive attribute and nothing else, and we show that, under the stated assumptions, perfect reconstruction together with population-level independence is sufficient for perfect nonlinear guardedness, with preservation guarantees in the imperfect-reconstruction regime. Empirically, we attain a superior utility-erasure trade-off across multiple datasets, sensitive attributes, and language models, surpassing even the supervised variants of competing methods despite using no target-task labels, while being simpler to optimize and converging substantially faster. Beyond erasure, our framework yields concept conversion (representation steering) as a by-product: by changing only the target concept at inference time, it produces representations corresponding to different concept values - at no additional training cost - a capability no prior erasure method provides, and on which we likewise demonstrate strong empirical performance.
Oct 23, 2026

Concept Acupuncturist: An Efficient and Precise Concept Erasure Method for Diffusion Models via Sparse Autoencoders

Diffusion-based image generation models produce high-quality images but may generate harmful or policy-violating content, motivating machine unlearning (MU) to selectively remove target concepts while preserving general capabilities. However, existing MU approaches often struggle with the distributed and entangled nature of concept representations, leading to inefficient optimization and imprecise edits when applied through large-scale fine-tuning over full diffusion trajectories. To bridge this gap, we propose Concept Acupuncturist (ConceptACP), a novel MU framework that integrates interpretability-guided analysis with inference-free fine-tuning for efficient and precise concept removal. By establishing mappings between semantic concepts and cross-attention blocks in the U-Net denoiser and localizing the most concept-relevant block for targeted optimization, ConceptACP avoids full diffusion trajectory optimization, thereby improving fine-tuning efficiency. To mitigate concept entanglement, we propose a feature-level correction mechanism based on Sparse Autoencoders (SAEs), where SAE-derived sparse, semantically meaningful feature activations are manipulated to construct targeted correction signals, which are then projected back to the output space of the corresponding block and used as supervision for precise concept removal. Consequently, ConceptACP eliminates the need for external inference-time modules while achieving efficient and fine-grained unlearning. Instantiated on SDXL-Turbo, ConceptACP improves the forgetting success rate by 36.34 percentage points in the standard one-step generation setting on average over the best-performing baseline, while preserving image quality and prompt fidelity, and generalizes effectively across other Stable Diffusion backbones. Our code is available at https://github.com/Eran-Xia/ConceptACP.
Oct 20, 2026

Push-Pull Attentional Anchoring for Diffusion Concept Erasure

Rapid advances in diffusion models have raised concerns about privacy, harmful content, and copyright violations. Concept erasure methods aim to address these risks by removing undesirable concepts from pretrained models without full retraining. However, existing approaches often degrade non-target generation quality through heuristic strength scaling (e.g., classifier-free guidance or negative prompt scaling), leading to uncontrolled and excessive semantic drift. To address this, we propose Push-Pull Attentional Anchoring (PPAA), a mechanism in the cross-attention space that displaces target representations while anchoring the scale of erasure via bounded cosine-similarity constraints. By using a relative similarity ratio instead of an absolute difference, our adaptive threshold represents a meaningful percentage of semantic retention, facilitating a single value to be applied uniformly across layers. We conduct extensive experiments across multiple concept categories, including celebrities, artistic styles, nudity, and intellectual property. Our method achieves favorable trade-offs against prior approaches on Stable Diffusion v1.4 and demonstrates its applicability to modern architectures such as FLUX.
Oct 9, 2026

Shared Canonical Component Removal Beyond Expert Views

In clinical practice, experts often begin by converting raw inputs into a set of expert features before making a decision. While black-box models trained on raw input can outperform models trained on expert views, it remains unclear what predictive signal the raw model exploits beyond what the expert view captures. We introduce an audit framework to quantify this gap using Shared Canonical Component Removal (SCCR). Unlike standard residualization or concept erasure, SCCR selectively removes only those directions strongly shared between learned raw and expert embeddings. Evaluations on synthetic data, coronary artery disease (CAD) detection, antidepressant treatment dropout prediction, and skin lesion classification using SkinCon annotations show that SCCR identifies the ways in which expert views are insufficient and recovers residual predictive signal while minimizing redundancy with the expert view.
Oct 6, 2026

Detecting a Shift Is Not Enough: Exact Minimax Limits of Linear Representation Repair

A mean shift between two data sources can be easy to detect but hard to remove without substantially changing their representations. We cast its removal as a statistical decision problem: from noisy differences between paired calibration measurements in Rd\mathbb{R}^d, learn one linear map, applied to both sources under a hard distortion budget, that leaves as little of the shift as possible on fresh data. We derive the exact finite-sample minimax risk over all such maps, (d−k)E[1/(d+2J)](d-k) \mathbb{E}[1/(d+2J)] with J∼Pois(κ/2)J\sim\mathrm{Pois}(κ/2), where the budget allows deleting kk directions and κκ is the calibration signal-to-noise ratio. Projecting out the mean calibration difference attains it without knowing κκ or the noise scale. This exposes a detection-repair gap: detecting the shift needs only κ≫dκ\gg\sqrt d, whereas removing a fixed fraction of it at constant distortion needs κ≍dκ\asymp d, as for estimating its direction. Standard linear concept erasers (MP, SAL, LEACE) remove the same calibration difference, so the formula gives, before fitting, exactly how much shift they leave on fresh data and how much calibration a target requires. The limit is robust: pairing keeps it exact for non-Gaussian shared content, the projection keeps its guarantee under anisotropic noise, and selective abstention cannot close the gap. On paired clinical and wearable sleep EEG, where differences between participants act as calibration noise, the formula predicts the device shift left in new participants, and more recordings per person soon stop helping. Together, these results tell whether a correction that falls short needs a better method, more recordings, or more participants.
Oct 5, 2026

Compositional Concept Erasure in Text-to-Image Diffusion Models via Hierarchically Grounded Semantic Surgery

Removing copyrighted, unsafe, or user-specified concepts from a deployed text-to-image diffusion model is now a practical requirement. Weight-editing methods can suppress fixed targets, but they require per-target retraining and modify the model checkpoint. Training-free methods, on the other hand, are deployment-friendly, but they suffer from text-side routing failures on compositional prompts. In such prompts, the erase target may be invoked through a related class rather than its lexical name, and its modifiers may migrate onto preserved objects. This paper proposes Hierarchically Grounded Semantic Surgery (HGSS), a training-free framework for compositional concept erasure. The framework lifts both the routing signal and the edit operator used by text-side erasure. First, hierarchical span grounding resolves erase-target spans through lexical, taxonomic, and semantic evidence, while guarding against broad-hypernym and compound-head false positives. Second, dynamic attribute binding refines the text conditioning during early denoising via a counterfactual reference and a preserve-aware cross-attention objective, keeping surviving attribute-noun bindings intact. HGSS selectively removes the erase target without updating model weights or adding learned parameters. On SEE, HGSS cuts hierarchical evasion from 29.54 to 10.02 and roughly halves pairwise attribute leakage, achieving the best Neighbor E and AttrP scores among the reported erasure methods. On UnlearnCanvas, HGSS slightly improves the six-metric average over the matched Semantic Surgery baseline, reaching state-of-the-art.
Oct 4, 2026

Your Unlearning Gives You Away: Identifying Erased Concepts in Diffusion Models

Existing attacks on unlearned diffusion models assume that the erased concepts are known in advance and focus on recovering them. In practice, however, model providers may not disclose which concepts have been removed, and even with access to the original base model, an adversary may still lack a clear target to attack. In this paper, we aim to answer the following critical but overlooked questions: which concepts have been erased from the model, and how many have been erased in total? To this end, we present Tracer, a framework that rapidly and accurately identifies erased concepts and estimates their number. Tracer efficiently identifies erased concepts without generating and classifying images. By combining lightweight spectral analysis of weight footprints, it enables efficient search over large candidate vocabularies. To distinguish multiple erased concepts, we introduce a footprint coverage objective that guides sequential discovery. Tracer estimates the number of erased concepts by detecting a sharp decline in candidate confidence as the selected concepts account for the erasure footprint, without requiring labeled examples for calibration. The framework requires only lightweight linear algebra and limited forward probes, with no prior knowledge of the unlearning algorithm. Experiments across text-to-image and text-to-video backbones and diverse unlearning methods demonstrate that Tracer identifies erased concepts and estimates their number in seconds, achieving 150 to 137,000 times and 133 to 20,000 times speedups over MIA and brute-force search on image and video models, respectively, with substantially higher identification accuracy.
Sep 27, 2026

Concept Score Relearning: A Unified Cross-Architecture Attack on Concept Erasure

Concept erasure aims to suppress undesirable knowledge in text-to-image generative models. However, existing robustness evaluations typically rely on relearning attacks tailored to specific model architectures. We study concept reactivation across two substantially different generative paradigms: noise-prediction U-Nets and flow-matching Transformers. We introduce Concept Score Relearning (CSR), a unified parameter-level framework that reactivates erased concepts by optimizing each model within its native prediction space. CSR requires no external target-concept image dataset and applies the same concept-directed objective to both U-Net-based Stable Diffusion and Transformer-based FLUX. Experiments across diverse concepts and multiple erasure methods demonstrate consistent concept reactivation across both architectures, highlighting the cross-architecture applicability of CSR and the persistent recoverability of apparently erased concepts. For strict nudity, CSR reaches average ASRs of 50.47% on FLUX and 40.29% on Stable Diffusion, consistently ranking first across all evaluated safety settings.
Sep 23, 2026

Hidden not Deleted: How Networks Suppress Entangled Features

When a neural network stops reporting a feature, has it forgotten the feature or merely hidden it? Shared representations complicate the answer: removing one feature's direction can also damage another feature. We study this problem in a small autoencoder that compresses and reconstructs inputs, allowing individual feature contributions to be traced directly. With the representation fixed, restoring a single bias partially recovers target detection by exposing a signal hidden behind the output's activation threshold. Tests on target-absent inputs distinguish this recovery from interventions that simply increase output. The same bias intervention becomes less effective when the representation is allowed to change. We find that training only the other outputs can weaken the target's contribution, even without a loss on its own output. A conditional theoretical analysis explains how interference with retained predictions creates this pressure. Yet more flexible predictors still detect residual target information. These findings show how recovery depends on feature geometry and on which parameters are trained, and why a silent output is insufficient evidence of forgetting.
Sep 11, 2026

Privacy-Preserving Deep Joint Source-Channel Coding with In-Loop Concept Erasure

Deep joint source-channel coding (DeepJSCC) transmits learned semantic features efficiently but can leak sensitive attributes such as gender, race, or speaker identity. We propose LEAPSC (LEACE-in-the-loop privacy for semantic communication), whose core contribution is the integration of in-loop least-squares concept erasure (LEACE) within a variational information bottleneck (VIB) encoder. By periodically refitting the projection operator during training, LEAPSC couples the encoder dynamics to the erasure mechanism, driving attribute-conditional mean differences toward zero within each task-label group on the fitting sample. Additional components, namely conditional value-at-risk (CVaR) tail-sensitive privacy, feature-wise linear modulation (FiLM) signal-to-noise ratio conditioning, and Lagrangian dual ascent, improve robustness across channel conditions and over the high-leakage tail of samples. On CelebA, FairFace, and Google Speech Commands, LEAPSC reaches task accuracy of 0.862, 0.755, and 0.925 respectively, with attacker accuracy at or below the label-only floor on CelebA (0.548 vs. floor 0.580) and within 2 percentage points (pp) of chance elsewhere, improving over an information-bottleneck adversarial baseline (IBAL) at a matched 52-epoch budget by +3.6, +2.5, and +1.3 pp (Welch's t-test, p=0.019 on CelebA).
Sep 11, 2026

GRACE: Adaptive Concept Erasure with Geometry-Guided Retention in Diffusion Models

Text-to-image (T2I) diffusion models inevitably internalize sensitive or non-compliant concepts from large-scale pretraining data, necessitating post-hoc concept erasure. However, existing erasure methods often lack explicit constraints on parameter updates, leading to over-intervention and unintended semantic drift. In addition, many methods rely on manually crafted counterfactual supervision, such as surrogate prompts, which incurs substantial data construction costs that limit scalability to new concepts. To address these limitations, we propose GRACE, a structured concept erasure framework designed to enable localized and selective intervention. Specifically, we introduce a semantically weighted sensitive subspace estimation to precisely lock intervention directions, and employ lightweight subspace-constrained adapters to prevent global semantic disturbance. To eliminate the dependency on manual prompt engineering, we design an automatically decoupled safe-anchor mechanism. To mitigate semantic drift induced by excessive intervention, we introduce an energy-driven dynamic gating mechanism that adaptively controls the timing and strength of intervention at inference. Extensive experiments demonstrate that our method achieves a superior balance between erasure effectiveness and generation fidelity. Compared with the average performance of five state-of-the-art (SOTA) concept erasure methods, our method improves the fine-grained NSFW reduction rate by 17.86%17.86\%, while reducing the macro-averaged target CLIP Score and preservation-oriented Fréchet Inception Distance (FID) by 4.55%4.55\% and 52.14%52.14\%, indicating stronger concept suppression with substantially improved preservation of the original model's generative utility.
Sep 10, 2026

Certifying Concept Unlearning in Text-to-Image Diffusion Models

Existing evaluations of concept unlearning in text-to-image (T2I) diffusion models primarily rely on attack success rates obtained through automated adversarial prompt search. However, these metrics provide only empirical evidence over a finite set of queries and leave residual leakage over the broader prompt space largely unquantified. This limitation can lead to overestimating unlearning effectiveness and underestimating safety risks. To address this gap, we introduce a novel certification framework for T2I concept unlearning that provides high-confidence guarantees with bounded error on residual concept leakage. Our approach combines statistical certification with worst-case analysis along concept-relevant embedding directions to derive explicit upper bounds on leakage probability under user-specified confidence levels. We evaluate our framework across three major concept categories namely NSFW content, artistic styles, and celebrity identities, and six state-of-the-art unlearning methods. Certified leakage bounds consistently exceed standard attack success rates by 16.2%, uncovering substantial residual risks missed by existing evaluation protocols. Crucially, our results demonstrate that empirical attack-based evaluations can significantly underestimate residual leakage and establish certification as a necessary complement for reliable auditing of concept unlearning in T2I diffusion models.
Sep 10, 2026

MUtE: A Dual Framework for Concept Erasure and Counterfactual Interventions

Erasing concept-specific information from representations has been proven useful for mitigating bias or interpreting model decisions. The joint objective is to transform the original representations such that the target concept becomes unpredictable, while maximally preserving concept-unrelated information. In this work, we revisit the optimal bounds of concept erasure to derive a novel class of erasure functions that naturally induce a deterministic, dual counterfactual mapping. Bridging the gap between theoretical optimality and practical representation learning, we design an implementation that imposes a translational bias on counterfactual trajectories - a constraint that aligns with how many concepts geometrically manifest in modern language models. Our framework enables seamless navigation between concept erasure and counterfactual generation. We empirically demonstrate its efficacy in improving downstream algorithmic fairness and generating counterfactual texts.
Aug 13, 2026

Semantic Steering for Controllable Generation: Tuning-Free Concept Erasure in Multimodal Diffusion Transformers

Multimodal Diffusion Transformers (MM-DiTs) have demonstrated remarkable text-to-image generation performance, surpassing traditional U-Net-based diffusion models. Nevertheless, their powerful generative capabilities also raise significant safety concerns, as they may generate sensitive or inappropriate content. While existing concept erasure methods aim to mitigate such risks, most require modifying model parameters, which are often architecture-specific and impractical for deployed larger models. Several tuning-free approaches face challenges when applied to advanced large-scale MM-DiTs due to their deeply embedded knowledge, broad semantic space, and context-dependent text encoders. To address these challenges, we propose to erase concepts by directly manipulating the model's internal representations. Our key insight, derived from an in-depth analysis of MM-DiT's block-wise generative roles, is that text-conditioned semantic representations are most salient in the middle blocks of MM-DiTs. Based on this, we extract representations of an unwanted concept and a desirable safe one from the middle block, construct a steering vector from their difference, and inject this single vector into consecutive early and middle blocks. By operating exclusively on the sparse text-branch tokens and leveraging the straight sampling trajectory of rectified flow, our method achieves effective concept erasure with negligible overhead and without any training. Extensive experiments across MM-DiT models demonstrate that our method achieves state-of-the-art performance in erasing diverse concepts, enables effective control over the final output, and remains robust to adversarial attacks.
Aug 11, 2026

Iterative Erasure Count Is Not an Affine-Invariant Concept Dimension

How many directions does a neural representation use to encode a concept? A common answer repeatedly erases probe directions and reports the stopping count or cumulative removed rank. We show that both quantities can change under an information-preserving invertible reparameterization, so neither is intrinsically a concept dimension. We distinguish model-defined population quantities (generating dimension, sufficient linear dimension, and minimum guarding rank) from procedure-defined quantities such as stopping count and cumulative edit rank. In a population Gaussian construction, an invertible shear preserves the prediction problem and all three quantities, yet changes the cumulative Euclidean erasure count from one to two. The separation holds for Moore--Penrose ordinary least squares and every finite nonnegative ridge weight. For a two-output full-QR procedure matching our motivating video analysis, cumulative edit rank similarly changes from two to the ambient dimension four. Conversely, the complete cumulative metric-QR trajectory is affine-equivariant when its positive-definite metric, probe, regularizer, and tie-breaking are transported consistently; exact covariance is one corollary, not a canonical semantic metric. In a known-rank finite-sample Adam/QR calibration, identity mixing stops after one accepted update in all 20 large-sample runs, whereas each tested shear a∈{.5,.75,1,1.25,2}a\in\{.5,.75,1,1.25,2\} accepts at least two updates in all 20 runs. Controlled reparameterizations of frozen V-JEPA2 features preserve rank-zero predictions yet alter later Euclidean trajectories under practical optimization. These visual contact experiments are stress tests, not estimates of contact dimension. Iterative erasure therefore returns a procedure-relative estimand jointly determined by representation geometry and the full measurement procedure, not a semantic dimension by itself.
Aug 7, 2026

FlowErase-OPD: Multi-Concept Erasure via Anchored On-Policy Distillation in Flow Matching Models

Recent advances in flow matching models have substantially improved the quality of text-to-image generation, but have also raised increasing safety concerns due to their potential to generate harmful or undesirable content. Existing concept erasure methods for flow matching models predominantly focus on removing individual concepts, while effectively erasing multiple concepts simultaneously remains challenging. We propose FlowErase-OPD, a framework for multi-concept erasure based on on-policy distillation (OPD). Our approach first distills multiple single-concept erased models into a unified LoRA module and introduces Anchored Multi-Teacher Distillation (AMTD), which incorporates a retention teacher to mitigate the trade-off between concept erasure and preservation of generative capabilities. To further improve the coordination of multiple erasure objectives, we develop Adaptive Retention Control (ARC), which dynamically adjusts the sampling frequency and loss weight of each erasure teacher, together with the relative contribution of erasure and retention teachers throughout training. Extensive experiments on nudity, object, and artistic-style erasure demonstrate that FlowErase-OPD consistently improves the trade-off between erasure effectiveness, image quality, and semantic alignment, achieving state-of-the-art performance across diverse multi-concept erasure settings. Furthermore, the resulting models exhibit strong robustness against adversarial attacks. These results highlight the potential of on-policy distillation as a principled framework for safe and controllable generation in flow matching models.
Aug 7, 2026

Are Visual Place Recognition Models Recognizing Places or Conditions? Distractor-Augmented Evaluation and Condition Suppression

Long-term Visual Place Recognition (VPR) is typically evaluated by matching queries from one condition against a database from another. Crowdsourced map databases, however, may mix conditions and include images that resemble the query in condition but depict different places. In the presence of these distractors, a method may retrieve by condition similarity rather than place identity. We argue that this susceptibility arises because the discriminability of VPR methods allows them to encode information such as illumination, weather, and seasonal appearance in their descriptors. We therefore introduce Distractor-Augmented Recall (DAR) to isolate and quantify the effect of distractors, and propose condition suppression to remove condition information from VPR descriptors. Across eleven methods and six datasets, method rankings under DAR@1 differ from those under Recall@1 (R@1), while applying INLP and LEACE as condition suppression methods generally improves DAR@1 without reducing R@1. Thus, distractor robustness is distinct from standard retrieval performance and can be improved by suppressing condition information.
Jul 29, 2026

CleanVideo: Adaptive Concept Erasure for Text-to-Video Diffusion Models

Concept erasure aims to selectively eliminate undesired visual semantics from pre-trained generative models without compromising their general utility. Extending concept erasure from images to video is nontrivial. Target concepts emerge gradually and vary across frames and denoising steps. As a result, fixed interventions may miss the target or introduce blurring, jitter, and content distortion. We propose CleanVideo, a selective erasure framework that performs low-dimensional subspace intervention controlled by a tri-modal gating mechanism. By jointly processing spatiotemporal visual features, timestep signals, and textual semantics, CleanVideo determines where, when, and whether to intervene, steering erased content toward natural surrogate concepts when such surrogates can be clearly defined while preserving non-target content. Experiments on three video diffusion models show that CleanVideo effectively erases target concepts while maintaining visual fidelity and temporal coherence, outperforming existing baselines under frame-level and video-level evaluations and under concept-recovery attacks when the protected pipeline remains intact.
Jul 27, 2026

LU-500: A Logo Benchmark for Concept Unlearning

Concept unlearning is increasingly used to limit the reproduction of protected or unsafe visual concepts in text-to-image models. Existing evaluations, however, mostly study targets that dominate the whole image, such as styles, broad object categories, or portrait-like identities, leaving company logos comparatively underexamined. Logos create a different failure mode: a small localized mark can carry the entire protected concept, must be visually precise to remain recognizable, and can be triggered implicitly by products, storefronts, packaging, or advertisements even when the word ``logo'' is absent. We introduce LU-500, a logo-unlearning benchmark built from Fortune Global 500 companies to study this localized and semantically entangled setting. LU-500 contains nearly 10,000 curated text-query and logo-image pairs, with an explicit track (LUex-500) and an implicit contextual track (LUim-500). To avoid reducing the task to a binary detector score, we define a multi-grained protocol that evaluates both local logo removal and global image preservation in pixel and latent spaces. Experiments on representative inference-time methods, including NP, SLD, and SEGA, and compatible fine-tuning-based methods such as ESD and Forget-Me-Not, show that the evaluated methods struggle to remove logo evidence without changing non-target content. We further analyze ProLU, a prompt-space multi-agent baseline: it improves local erasure by removing logo-inducing semantics, but also illustrates why prompt filtering is not a substitute for weight-level disentanglement. Correlation analyses over logo area, location, and structural complexity suggest that future logo unlearning may need spatially aware controls, such as SSIM-guided constraints, rather than purely global concept suppression.
Jul 26, 2026

To Erase, or Not to Erase: Robust Training-Free Concept Erasure with Preservation aware Adaptive Ranked Subspace Expansion

Concept erasure techniques (CETs) edit text-to-image diffusion models to erase undesired targets such as NSFW content or copyrighted styles, while preserving model utility on benign concepts. Current CETs face a trade-off between erasure robustness and utility: stronger edits erase the target more reliably but degrade utility on non-target concepts, and vice versa. This stems from how existing methods define what to erase and what to preserve. Many CETs rely on static concept banks specified manually, generated by LLMs, or selected by CLIP image-text similarity. Such banks do not model how prompts steer the model during denoising, leaving it vulnerable to triggers that reintroduce the target while suppressing nearby benign concepts. We present Preservation-aware Adaptive Ranked Subspace Expansion (PARSE), a training-free framework for robust concept erasure in latent diffusion models. Given a target, PARSE queries the diffusion model with classifier-free guidance to dynamically discover target-inducing erase concepts and nearby retain concepts in the model vocabulary. It then edits the cross-attention value space with a preservation-aware projection that removes target directions while leaving retain directions intact. For triggers beyond this vocabulary-indexed space, PARSE iteratively searches for re-emergence triggers by textual inversion and adaptively expands the erased subspace only when a new trigger direction does not conflict with retain semantics. We also introduce the Balanced Erasure Utility Score (BEUS), which combines robustness (ASR under multiple attacks) and utility preservation (FID) via bounded monotone transforms and harmonic mean aggregation. Experiments on NSFW, artistic style, and object erasure, with a large-scale robustness-utility analysis over many CET baselines, show that PARSE erases multiple concepts robustly without sacrificing post-edit utility.
Jul 20, 2026

Stress Testing Concept Erasure with Large Language Model Agents

Concept erasure aims to remove semantic concepts from a trained generative model and is increasingly important for responsible AI deployment. However, verifying whether a model has robustly removed targeted concepts remains a critical challenge. Existing evaluation methods are typically pre-defined and static, failing to expose vulnerabilities under diverse natural-language probes and challenging conditions. Moreover, manually designed evaluation strategies can be biased and difficult to scale. We posit that concept erasure evaluation is best formulated as an adaptive hypothesis search, operationalised by agents that iteratively propose, critique, and verify tests to systematically expand coverage of failure modes. To this end, we propose Stress Testing Agents for Concept Erasure (STACE), a framework that autonomously stress-tests concept-erased models using multiple Large Language Model (LLM) agents, by iteratively generating and verifying stress-testing hypotheses grounded by external knowledge. We also introduce a suite of metrics for assessing the performance and efficiency of LLM-agent-powered stress-testing frameworks. Our extensive experiments show that STACE outperforms five LLM-based evaluation baselines on four concept categories. Further analysis across two T2I models, six concept erasure approaches, and various erasure strengths show that STACE is robust for different settings. We also show that STACE can be adapted beyond concept erasure evaluation to other problem domains, such as LLM jailbreaking. Our code is available anonymously.
Jul 16, 2026

Uni-AdaVD: Universal Concept Erasure for Visual Generation via Orthogonal Value Decomposition

Visual generative models inevitably absorb undesirable concepts from uncurated pretraining data, making concept erasure essential for safe deployment. Existing erasure methods, however, are often architecture-specific and struggle to remove target concepts while preserving non-target content and generative priors. We present Uni-AdaVD, a universal inference-time concept erasure framework for visual generation. Uni-AdaVD treats the value space of multimodal attention as a unified intervention space and introduces encoder-aware target representation construction to localize target semantics across heterogeneous text encoders. It further combines orthogonal value decomposition with an adaptive erasing shift to suppress target semantic directions without updating the original model weights. Extensive experiments on U-Net-, DiT-, and autoregressive image generators, as well as text-to-video models, demonstrate strong performance on single- and multi-concept erasure while preserving non-target priors. These results suggest that Uni-AdaVD provides an efficient and adaptable safety mechanism for modern visual generative models. Our code is available at https://github.com/QifanZhou/Uni-AdaVD.
Jul 7, 2026

TILDE: TILt-based Distributional Erasure for Concept Unlearning

Concept unlearning in text-to-image diffusion models is critical for safe and practical deployment: with rising privacy concerns, copyright disputes, trademark constraints, and safety regulations, deployed systems must be able to suppress unwanted concepts after training. Existing methods often remove the target concept effectively, but practical unlearning also requires an equally fundamental property: the unlearned model should retain quality, diversity, and semantic coverage on benign generation. The gold standard is a retain-only model trained from scratch without the unwanted data. However, common erasure objectives do not specify which post-unlearning distribution should approximate this reference, leaving retention as an implicit consequence of the update rule. We propose TILDE, TILt-based Distributional Erasure, which formulates concept unlearning as a distributional alignment problem: the desired target is the minimum-deviation conditional distribution from the pretrained model under a forgetting constraint. This energy-tilted, anchor-free target suppresses concept-expressing images while preserving benign relative mass for each prompt. We instantiate this principle with residual ∇\nabla-GFlowNet training, which learns the score correction induced by the forget energy relative to the pretrained diffusion model. Across objects, artistic styles, and characters, TILDE achieves strong forgetting while improving retention and distributional fidelity over prior baselines.
Jul 6, 2026

Erasing Without Collateral Damage: Precise Concept Removal in Diffusion Models

Training-free concept erasure is an attractive mechanism for controlling text-to-image diffusion models, but precise erasure often comes at the cost of damaging semantically related non-target concepts. Existing value-space methods remove the component of each cross-attention value along the target concept direction, implicitly treating target identity and shared visual structure as the same signal. We argue that this is the source of much of the collateral damage in prior preservation. We introduce CARE, a closed-form concept erasure operator that replaces the raw target direction with a kept-subspace-aware direction computed from a small bank of retained concept anchors. The resulting edit is applied directly in cross-attention value space, requires no model fine-tuning, and adds only a negligible offline computation. A single shrinkage parameter controls the erase-preserve trade-off. We further show that the operator admits a minimum-disturbance interpretation and, in its projection form, leaves the kept subspace invariant. Experiments under the standard concept-erasure protocol show that our method preserves non-target concepts more faithfully while maintaining competitive erasure across instance, style, and celebrity concepts. Code: https://github.com/parthupman/care