Diffusion Model Unlearning

Follow this topic Ask about this topic

Diffusion model unlearning studies how to suppress unwanted content—such as an identity, style, object, or unsafe imagery—in image and video generators without broadly damaging their abilities. Methods include fine-tuning and distillation, localized edits to weights or attention, and training-free interventions that steer activations, embeddings, or latents during generation. Evaluation increasingly tests not only whether prompts produce the target, but also adversarial recovery, related-concept retention, compositional quality, and persistence across frames or sequential edits. A central tension remains: concepts share representations, so stronger or more robust erasure can harm neighboring concepts, while apparent suppression may leave visual knowledge recoverable through other pathways.

Latest papers

77 papers19 papers in the last four weeks, up 138% on the four weeks before. 0.1% of all new papers.

A Concept is More Than a Word: Diversified Unlearning in Text-to-Image Diffusion Models
Mar 19, 2026Duc Hao Pham, Van Duy Truong, Duy Khanh Dinh +3Diffusion Model Unlearning Concept Erasure
VNPT AI, VNPT Group Hanoi, Vietnam · Independent Researcher Australia
TINA: Text-Free Inversion Attack for Unlearned Text-to-Image Diffusion Models
Mar 18, 2026Qianlong Xiang, Miao Zhang, Haoyu Zhang +3T2I Diffusion Models Diffusion Model Unlearning
Harbin Institute of Technology (Shenzhen) · City University of Hong Kong · Shenzhen Loop Area Institute +2
REFORGE: Multi-modal Attacks Reveal Vulnerable Concept Unlearning in Image Generation Models
Mar 17, 2026Yong Zou, Haoran Li, Fanxiao Li +5Adversarial Attacks Adversarial Robustness
Yunnan University · Northeastern University · National University of Singapore
RAZOR: Ratio-Aware Layer Editing for Targeted Unlearning in Vision Transformers and Diffusion Models
Mar 16, 2026Ravi Ranjan, Utkarsh Grover, Xiaomin Lin +1Machine Unlearning VLM Unlearning
Florida International University Miami, USA · University of South Florida Tampa, USA
Unlearning the Unpromptable: Prompt-free Instance Unlearning in Diffusion Models
Mar 11, 2026Kyungryeol Lee, Kyeonghyun Lee, Seongmin Hong +2Diffusion Model Unlearning Machine Unlearning
Department of Electrical and Computer Engineering (ECE), Seoul National University (SNU), Seoul, 08826, Republic of Korea · Institute of New Media and Communications (INMC), SNU, Seoul, 08826, Republic of Korea · ECE, INMC, IPAI, SNU, Seoul, 08826, Republic of Korea
Prototype-Guided Concept Erasure in Diffusion Models
Mar 9, 2026Yuze Cai, Jiahao Lu, Hongxiang Shi +2Concept Erasure Safe T2I Generation
Fudan University · National University of Singapore
Compensation-free Machine Unlearning in Text-to-Image Diffusion Models by Eliminating the Mutual Information
Mar 1, 2026Xinwen Cheng, Jingyuan Zhang, Zhehao Huang +2Diffusion Model Unlearning Concept Erasure
Institute of Image Processing and Pattern Recognition, School of Automation and Intelligent Sensing, Shanghai Jiao Tong University
EraseAnything++: Enabling Concept Erasure in Rectified Flow Transformers Leveraging Multi-Object Optimization
Mar 1, 2026Zhaoxin Fan, Nanxiang Jiang, Daiheng Gao +2Concept Erasure Diffusion Model Unlearning
Beijing Advanced Innovation Center for Future Blockchain and Privacy Computing, School of Artificial Intelligence, Beihang University. · University of Science and Technology of China.
Forgetting is Competition: Rethinking Unlearning as Representation Interference in Diffusion Models
Mar 1, 2026Ashutosh Ranjan, Vivek Srivastava, Shirish Karande +1Diffusion Model Unlearning Machine Unlearning
TCS Research, India · Kalinga Institute of Industrial Technology
Roots Beneath the Cut: Uncovering the Risk of Concept Revival in Pruning-Based Unlearning for Diffusion Models
Feb 28, 2026Ci Zhang, Zhaojun Ding, Chence Yang +7Diffusion Model Unlearning Machine Unlearning
University of Georgia · Carnegie Mellon University · Northeastern University +2
Localized Concept Erasure in Text-to-Image Diffusion Models via High-Level Representation Misdirection
Feb 23, 2026Uichan Lee, Jeonghyeon Kim, Sangheum HwangDiffusion Model Unlearning T2I Diffusion Models
Department of Data Science, Seoul National University of Science and Technology
Selective Fine-Tuning for Targeted and Robust Concept Unlearning
Feb 8, 2026Mansi, Avinash Kori, Francesca Toni +1Fine-Tuning Adversarial Robustness
Department of Computing Imperial College London, United Kingdom
Locality-Aware Continual Unlearning for Diffusion Models
Dec 2, 2025Naveen George, Naoki Murata, Yuhta Takida +2Diffusion Model Unlearning Machine Unlearning
Indian Institute of Technology Hyderabad, India · Sony AI, Japan · Sony Group Corporation, Japan
Rethinking Robust Adversarial Concept Erasure in Diffusion Models
Oct 31, 2025Qinghong Yin, Yu Tian, Heming Yang +5Adversarial Robustness Diffusion Model Unlearning
Beijing University of Posts and Telecommunications, China · University of Chinese Academy of Sciences, China · Nanjing University of Aeronautics and Astronautics, China
Not Every Time and Frequency Need to Be Forgotten in Diffusion Unlearning
Oct 20, 2025Jinseong Park, Mijung ParkDiffusion Model Unlearning Machine Unlearning
Korea Institute for Advanced Study · University of British Columbia
UNO: Unlearning via Orthogonalization in Generative models
Jun 5, 2025Pinak Mandal, Georg A. GottwaldMachine Unlearning Diffusion Model Unlearning
University of Sydney, Australia
The Linear Geometry of Interpretable Tokens: Jailbreaking Attacks and Defenses for Unlearned Diffusion Models
Apr 30, 2025Siyi Chen, Yimeng Zhang, Sijia Liu +1Jailbreak Attacks Diffusion Model Unlearning
University of Michigan · Michigan State University

Mar 19, 2026J/K move · Enter open · S save

A Concept is More Than a Word: Diversified Unlearning in Text-to-Image Diffusion Models

Duc Hao Pham, Van Duy Truong, Duy Khanh Dinh, Tien Cuong Nguyen+2

VNPT AI, VNPT Group Hanoi, Vietnam · Independent Researcher Australia

Diffusion Model Unlearning Concept Erasure Machine Unlearning T2I Diffusion Models

Open paper AskPDF arXiv

Concept unlearning has emerged as a promising direction for reducing the risks of harmful content generation in text-to-image diffusion models by selectively erasing undesirable concepts from a model's parameters. Existing approaches typically rely on keywords to identify the target concept to be unlearned. However, we show that this keyword-based formulation is inherently limited: a visual concept is multi-dimensional, can be expressed in diverse textual forms, and often overlap with related concepts in the latent space, making keyword-only unlearning, which imprecisely indicate the target concept is brittle and prone to over-forgetting. This occurs because a single keyword represents only a narrow point estimate of the concept, failing to cover its full semantic distribution and entangled variations in the latent space. To address this limitation, we propose Diversified Unlearning, a distributional framework that represents a concept through a set of contextually diverse prompts rather than a single keyword. This richer representation enables more precise and robust unlearning. Through extensive experiments across multiple benchmarks and state-of-the-art baselines, we demonstrate that integrating Diversified Unlearning as an add-on component into existing unlearning pipelines consistently achieves stronger erasure, better retention of unrelated concepts, and improved robustness against adversarial recovery attacks.