paper-with-me

Papers

Multi-concept Model Immunization through Differentiable Model Merging

2024-12-19 · Amber Yijia Zheng, Raymond A. Yeh

Model immunization is an emerging direction that aims to mitigate the potential risk of misuse associated with open-sourced models and advancing adaptation methods. The idea is to make the released models' weights difficult to fine-tune on certain harmful applications, hence the name `immunized''. Recent work on model immunization focuses on the single-concept setting. However, models need to be immunized against multiple concepts in real-world situations. To address this gap, we propose an immunization algorithm that, simultaneously, learns a single `difficult initialization'' for adaptation methods over a set of concepts. We achieve this by incorporating a differentiable merging layer that combines a set of model weights adapted over multiple concepts. In our experiments, we demonstrate the effectiveness of multi-concept immunization by generalizing prior work's experiment setup of re-learning and personalization adaptation to multiple concepts.

📄 PDF Abstract BibTeX arXiv:2412.15320

Code (1)

amberyzheng/MIMA 공식 구현 pytorch

Tasks

model

Methods 이 논문이 사용한 방법론

SET Dynamic Sparse Training method where weight mask is updated randomly periodically

Similar Papers 제목 키워드 기반

Graph Adversarial Immunization for Certifiable Robustness

2023-02-16 · Shuchang Tao, HuaWei Shen, Qi Cao, Yunfan Wu 외

Despite achieving great success, graph neural networks (GNNs) are vulnerable to adversarial attacks. Existing defenses focus on developing adversarial training or model modification. In this paper, we propose and formula…

Adversarial AttackCombinatorial Optimization

GIFT: Gradient-aware Immunization of diffusion models against malicious Fine-Tuning with safe concepts retention

2025-07-18 · Amro Abdalla, Ismail Shaheen, Dan DeGenaro, Rupayan Mallick 외 arxiv

We present GIFT: a {G}radient-aware {I}mmunization technique to defend diffusion models against malicious {F}ine-{T}uning while preserving their ability to generate safe content. Existing safety mechanisms like safety ch…

Immunization against harmful fine-tuning attacks

2024-02-26 · Domenic Rosati, Jan Wehner, Kai Williams, Łukasz Bartoszcze 외

Large Language Models (LLMs) are often trained with safety guards intended to prevent harmful text generation. However, such safety training can be removed by fine-tuning the LLM on harmful datasets. While this emerging …

Text Generation

Semantic Mismatch and Perceptual Degradation: A New Perspective on Image Editing Immunity

2025-12-16 · Shuai Dong, Jie Zhang, Guoying Zhao, Shiguang Shan 외 arxiv

Text-guided image editing via diffusion models, while powerful, raises significant concerns about misuse, motivating efforts to immunize images against unauthorized edits using imperceptible perturbations. Prevailing met…

Image Editing

Model Immunization from a Condition Number Perspective

2025-05-29 · Amber Yijia Zheng, Cedar Site Bai, Brian Bullins, Raymond A. Yeh

Model immunization aims to pre-train models that are difficult to fine-tune on harmful tasks while retaining their utility on other non-harmful tasks. Though prior work has shown empirical evidence for immunizing text-to…

model