paper-with-me

Papers

Safety Alignment via Constrained Knowledge Unlearning

2025-05-24 · Zesheng Shi, Yucheng Zhou, Jing Li

Despite significant progress in safety alignment, large language models (LLMs) remain susceptible to jailbreak attacks. Existing defense mechanisms have not fully deleted harmful knowledge in LLMs, which allows such attacks to bypass safeguards and produce harmful outputs. To address this challenge, we propose a novel safety alignment strategy, Constrained Knowledge Unlearning (CKU), which focuses on two primary objectives: knowledge localization and retention, and unlearning harmful knowledge. CKU works by scoring neurons in specific multilayer perceptron (MLP) layers to identify a subset U of neurons associated with useful knowledge. During the unlearning process, CKU prunes the gradients of neurons in U to preserve valuable knowledge while effectively mitigating harmful content. Experimental results demonstrate that CKU significantly enhances model safety without compromising overall performance, offering a superior balance between safety and utility compared to existing methods. Additionally, our analysis of neuron knowledge sensitivity across various MLP layers provides valuable insights into the mechanics of safety alignment and model knowledge editing.

📄 PDF Abstract BibTeX arXiv:2505.18588

Code (0)

등록된 구현이 없습니다.

Tasks

knowledge editingSafety Alignment

Similar Papers 제목 키워드 기반

Open Problems in Machine Unlearning for AI Safety

2025-01-09 · Fazl Barez, Tingchen Fu, Ameya Prabhu, Stephen Casper 외

As AI systems become more capable, widely deployed, and increasingly autonomous in critical areas such as cybersecurity, biological research, and healthcare, ensuring their safety and alignment with human values is param…

Machine Unlearning

SafeLLM: Unlearning Harmful Outputs from Large Language Models against Jailbreak Attacks

2025-08-21 · Xiangman Li, Xiaodong Wu, Qi Li, Jianbing Ni 외 arxiv

Jailbreak attacks pose a serious threat to the safety of Large Language Models (LLMs) by crafting adversarial prompts that bypass alignment mechanisms, causing the models to produce harmful, restricted, or biased content…

CATNIP: LLM Unlearning via Calibrated and Tokenized Negative Preference Alignment

2026-02-02 · Zhengbang Yang, Yisheng Zhong, Junyuan Hong, Zhuangdi Zhu arxiv

Pretrained knowledge memorized in LLMs raises critical concerns over safety and privacy, which has motivated LLM Unlearning as a technique for selectively removing the influences of undesirable knowledge. Existing approa…

CAP: Controllable Alignment Prompting for Unlearning in LLMs

2026-04-23 · Zhaokun Wang, Jinyu Guo, Jingwen Pu, Hongli Pu 외 arxiv

Large language models (LLMs) trained on unfiltered corpora inherently risk retaining sensitive information, necessitating selective knowledge unlearning for regulatory compliance and ethical safety. However, existing par…

Reinforcement Learning

SineProject: Machine Unlearning for Stable Vision Language Alignment

2025-11-23 · Arpit Garg, Hemanth Saratchandran, Simon Lucey arxiv

Multimodal Large Language Models (MLLMs) increasingly need to forget specific knowledge such as unsafe or private information without requiring full retraining. However, existing unlearning methods often disrupt vision l…