# Targeted Forgetting of Image Subgroups in CLIP Models

arXiv: https://arxiv.org/abs/2506.03117 · HTML: https://mlsanity.com/paper/2506.03117/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Placed in Machine unlearning, one of 12 directions in LLM safety and security.
- Builds on 42 of its 63 references that are in the map.
- First posted on arXiv 3 Jun 2025.

## Placed in

- LLM safety and security / [Machine unlearning](https://mlsanity.com/safety/machine-unlearning/)

## Builds on (most cited first, 30 of 42)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [LAION-5B: An open large-scale dataset for training next generation image-text models](https://mlsanity.com/paper/2210.08402/) (2022)
- [Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time](https://mlsanity.com/paper/2203.05482/) (2022)
- [Continual Lifelong Learning with Neural Networks: A Review](https://mlsanity.com/paper/1802.07569/) (2018)
- [Machine Unlearning](https://mlsanity.com/paper/1912.03817/) (2019)
- [Eternal Sunshine of the Spotless Net: Selective Forgetting in Deep Networks](https://mlsanity.com/paper/1911.04933/) (2019)
- [Who's Harry Potter? Approximate Unlearning in LLMs](https://mlsanity.com/paper/2310.02238/) (2023)
- [Remember What You Want to Forget: Algorithms for Machine Unlearning](https://mlsanity.com/paper/2103.03279/) (2021)
- [Unrolling SGD: Understanding Factors Influencing Machine Unlearning](https://mlsanity.com/paper/2109.13398/) (2021)
- [Machine Unlearning of Features and Labels](https://mlsanity.com/paper/2108.11577/) (2021)
- [Fast Yet Effective Machine Unlearning](https://mlsanity.com/paper/2111.08947/) (2021)
- [Model Sparsity Can Simplify Machine Unlearning](https://mlsanity.com/paper/2304.04934/) (2023)
- [Style Injection in Diffusion: A Training-free Approach for Adapting Large-scale Diffusion Models for Style Transfer](https://mlsanity.com/paper/2312.09008/) (2023)
- [Towards Fairer Datasets: Filtering and Balancing the Distribution of the People Subtree in the ImageNet Hierarchy](https://mlsanity.com/paper/1912.07726/) (2019)
- [BREEDS: Benchmarks for Subpopulation Shift](https://mlsanity.com/paper/2008.04859/) (2020)
- [Squeeze, Recover and Relabel: Dataset Condensation at ImageNet Scale From A New Perspective](https://mlsanity.com/paper/2306.13092/) (2023)
- [Zero-Shot Machine Unlearning](https://mlsanity.com/paper/2201.05629/) (2022)
- [Towards Foundation Models for Scientific Machine Learning: Characterizing Scaling and Transfer Behavior](https://mlsanity.com/paper/2306.00258/) (2023)
- [Machine Unlearning via Algorithmic Stability](https://mlsanity.com/paper/2102.13179/) (2021)
- [Variational Bayesian Unlearning](https://mlsanity.com/paper/2010.12883/) (2020)
- [CALIP: Zero-Shot Enhancement of CLIP with Parameter-free Attention](https://mlsanity.com/paper/2209.14169/) (2022)
- [A Whac-A-Mole Dilemma: Shortcuts Come in Multiples Where Mitigating One Amplifies Others](https://mlsanity.com/paper/2212.04825/) (2022)
- [Deep Unlearning via Randomized Conditionally Independent Hessians](https://mlsanity.com/paper/2204.07655/) (2022)
- [CLIP for All Things Zero-Shot Sketch-Based Image Retrieval, Fine-Grained or Not](https://mlsanity.com/paper/2303.13440/) (2023)
- [What to Pre-Train on? Efficient Intermediate Task Selection](https://mlsanity.com/paper/2104.08247/) (2021)
- [Domain Watermark: Effective and Harmless Dataset Copyright Protection is Closed at Hand](https://mlsanity.com/paper/2310.14942/) (2023)
- [Unified Gradient-Based Machine Unlearning with Remain Geometry Enhancement](https://mlsanity.com/paper/2409.19732/) (2024)
- [Evaluating Machine Unlearning via Epistemic Uncertainty](https://mlsanity.com/paper/2208.10836/) (2022)
- [Improved Fine-Tuning by Better Leveraging Pre-Training Data](https://mlsanity.com/paper/2111.12292/) (2021)

## Built on by (most cited first, 6 of 6)

- [SafeR-CLIP: Mitigating NSFW Content in Vision-Language Models While Preserving Pre-Trained Knowledge](https://mlsanity.com/paper/2511.16743/) (2025)
- [Models Designed to Forget: Machine Unlearning via Key Deletion](https://mlsanity.com/paper/2603.15033/) (2026)
- [CATA: Continual Machine Unlearning via Conflict-Averse Task Arithmetic](https://mlsanity.com/paper/2605.18610/) (2026)
- [Reference-Guided Machine Unlearning](https://mlsanity.com/paper/2603.11210/) (2026)
- [Why Instruction-Based Unlearning Fails in Diffusion Models?](https://mlsanity.com/paper/2604.01514/) (2026)
- [ICED: Concept-level Machine Unlearning via Interpretable Concept Decomposition](https://mlsanity.com/paper/2605.14309/) (2026)

## Abstract (arXiv)

Foundation models (FMs) such as CLIP have demonstrated impressive zero-shot performance across various tasks by leveraging large-scale, unsupervised pre-training. However, they often inherit harmful or unwanted knowledge from noisy internet-sourced datasets, compromising their reliability in real-world applications. Existing model unlearning methods either rely on access to pre-trained datasets or focus on coarse-grained unlearning (e.g., entire classes), leaving a critical gap for fine-grained unlearning. In this paper, we address the challenging scenario of selectively forgetting specific portions of knowledge within a class, without access to pre-trained data, while preserving the model's overall performance. We propose a novel three-stage approach that progressively unlearns targeted knowledge while mitigating over-forgetting. It consists of (1) a forgetting stage to fine-tune the CLIP on samples to be forgotten, (2) a reminding stage to restore performance on retained samples, and (3) a restoring stage to recover zero-shot capabilities using model souping. Additionally, we introduce knowledge distillation to handle the distribution disparity between forgetting, retaining samples, and unseen pre-trained data. Extensive experiments on CIFAR-10, ImageNet-1K, and style datasets demonstrate that our approach effectively unlearns specific subgroups while maintaining strong zero-shot performance on semantically similar subgroups and other categories, significantly outperforming baseline unlearning methods, which lose effectiveness under the CLIP unlearning setting.