Concept Erasure
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.
Look But Don't Touch with Sparse Autoencoders for Unlearning in Diffusion Models
Concept Removal Guidance: Evidence-Calibrated Negative Guidance for Safe Diffusion Sampling
ScaleErasure: Inference-Time Minimal Intervention for Precise Concept Erasure in Next-Scale Autoregressive Image Generation
Obliviate: Erasing Concepts from Autoregressive Image Generation Models
Co-occurring associated retained concepts in Diffusion Unlearning
SACE: Concept Erasure at the Semantic Singularity in Visual Autoregressive Models
On The Effectiveness-Fluency Trade-Off In LLM Conditioning: A Systematic Study
Debiasing Without Protected Attributes: Latent Concept Erasure from Textual Profiles
TRACER: Token ReAssignment for Concept ERasure in Generative Recommendation
Refit the Probe: Single-Direction Ablation Is Not a Necessity Test
Geometric Erasure by Contrastive Velocity Matching in Rectified Flows
Orthogonal Concept Erasure for Diffusion Models
Where Concept Erasure Should Occur: Concept-Layer Alignment in Text-to-Video Diffusion Models
Mosaic: Compositional Multi-Concept Erasure via Vector Field Blending
FlowErase-RL: Rethinking Concept Erasure as Reward Optimization in Flow Matching Models
Whispers in the Noise: Surrogate-Guided Concept Awakening via a Multi-Agent Framework
Empty SPACE: Cross-Attention Sparsity for Concept Erasure in Diffusion Models
SwordBench: Evaluating Orthogonality of Steering Image Representations
Probe-Geometry Alignment: Erasing the Cross-Sequence Memorization Signature Below Chance
A framework for analyzing concept representations in neural models
MidSteer: Optimal Affine Framework for Steering Generative Models
Closed-Form Concept Erasure via Double Projections
SHIFT: Steering Hidden Intermediates in Flow Transformers
We Should Distinguish Unlearning From Untraining
delete'' specific data points or behaviours from a trained model, a goal referred to as machine unlearning''. We argue that the umbrella term unlearning'' actually spans two distinct problem formulations, but the distinction between them has not yet been observed in literature. This causes ambiguity around when an unlearning algorithm is expected to work, leads to the use of inappropriate metrics and baselines when comparing algorithms to one another, difficulty in interpreting results, and missed opportunities for pursuing critical research directions. In this paper, we argue for the position that a fundamental distinction must be made between two notions that we refer to as unlearning and untraining. On one hand, untraining aims to reverse the effect of having trained on a given forget set'', i.e. to remove the influence that that specific set of examples had on the model during training. On the other hand, unlearning aims not just to remove the influence of those given examples, but also of the entire underlying subdistribution from which they were sampled (and thus e.g. the concept or model behaviour that those examples represent). We discuss technical definitions of these problems and map problem settings studied in the literature to each notion. By disambiguating technical definitions, our work aims to accelerate progress in this important field.Neighbor-Aware Localized Concept Erasure in Text-to-Image Diffusion Models
Z-Erase: Enabling Concept Erasure in Single-Stream Diffusion Transformers
A Concept is More Than a Word: Diversified Unlearning in Text-to-Image Diffusion Models
Prototype-Guided Concept Erasure in Diffusion Models
sexual'' or violent'', whose wide scope and multi-faceted nature make them difficult to erase reliably. To overcome this limitation, we exploit the model's intrinsic embedding geometry to identify latent embeddings that encode a given concept. By clustering these embeddings, we derive a set of concept prototypes that summarize the model's internal representations of the concept, and employ them as negative conditioning signals during inference to achieve precise and reliable erasure. Extensive experiments across multiple benchmarks show that our approach achieves substantially more reliable removal of broad concepts while preserving overall image quality, marking a step towards safer and more controllable image generation.