English
Related papers

Related papers: Gradient Surgery for Safe LLM Fine-Tuning

200 papers

Large Language Models (LLMs) exhibit impressive capabilities but also present risks such as biased content generation and privacy issues. One of the current alignment techniques includes principle-driven integration, but it faces challenges…

Computation and Language · Computer Science 2025-05-30 Yi Luo , Zhenghao Lin , Yuhao Zhang , Jiashuo Sun , Chen Lin , Chengjin Xu , Xiangdong Su , Yelong Shen , Jian Guo , Yeyun Gong

Fine-tuning large language models (LLMs) improves performance but introduces critical safety vulnerabilities: even minimal harmful data can severely compromise safety measures. We observe that perturbations orthogonal to the alignment…

Machine Learning · Computer Science 2026-01-08 Shuo Yang , Qihui Zhang , Yuyang Liu , Xiaojun Jia , Kunpeng Ning , Jiayu Yao , Jigang Wang , Hailiang Dai , Yibing Song , Li Yuan

Fine-tuning Large Language Models (LLMs) has emerged as a common practice for tailoring models to individual needs and preferences. The choice of datasets for fine-tuning can be diverse, introducing safety concerns regarding the potential…

Computation and Language · Computer Science 2024-10-15 Hyeong Kyu Choi , Xuefeng Du , Yixuan Li

Fine-tuning aligned language models on benign tasks unpredictably degrades safety guardrails, even when training data contains no harmful content and developers have no adversarial intent. We show that the prevailing explanation, that…

Although large language models (LLMs) achieve effective safety alignment at the time of release, they still face various safety challenges. A key issue is that fine-tuning often compromises the safety alignment of LLMs. To address this…

Computation and Language · Computer Science 2025-05-27 Di Wu , Xin Lu , Yanyan Zhao , Bing Qin

Safety alignment in large language models (LLMs) is achieved through fine-tuning mechanisms that regulate neuron activations to suppress harmful content. In this work, we propose a novel approach to induce disalignment by identifying and…

Machine Learning · Computer Science 2025-05-01 Yi Zhou , Wenpeng Xing , Dezhang Kong , Changting Lin , Meng Han

While deep learning and deep reinforcement learning (RL) systems have demonstrated impressive results in domains such as image classification, game playing, and robotic control, data efficiency remains a major challenge. Multi-task learning…

Machine Learning · Computer Science 2020-12-23 Tianhe Yu , Saurabh Kumar , Abhishek Gupta , Sergey Levine , Karol Hausman , Chelsea Finn

Optimizing large language models (LLMs) for downstream use cases often involves the customization of pre-trained LLMs through further fine-tuning. Meta's open release of Llama models and OpenAI's APIs for fine-tuning GPT-3.5 Turbo on custom…

Computation and Language · Computer Science 2023-10-06 Xiangyu Qi , Yi Zeng , Tinghao Xie , Pin-Yu Chen , Ruoxi Jia , Prateek Mittal , Peter Henderson

The current safeguard mechanisms for large language models (LLMs) are indeed susceptible to jailbreak attacks, making them inherently fragile. Even the process of fine-tuning on apparently benign data for downstream tasks can jeopardize…

Computation and Language · Computer Science 2024-05-16 Xin Yi , Shunfan Zheng , Linlin Wang , Xiaoling Wang , Liang He

The jailbreak attack can bypass the safety measures of a Large Language Model (LLM), generating harmful content. This misuse of LLM has led to negative societal consequences. Currently, there are two main approaches to address jailbreak…

Computation and Language · Computer Science 2024-03-25 Zezhong Wang , Fangkai Yang , Lu Wang , Pu Zhao , Hongru Wang , Liang Chen , Qingwei Lin , Kam-Fai Wong

Understanding and addressing potential safety alignment risks in large language models (LLMs) is critical for ensuring their safe and trustworthy deployment. In this paper, we highlight an insidious safety threat: a compromised LLM can…

Machine Learning · Computer Science 2026-03-24 Guangnian Wan , Xinyin Ma , Gongfan Fang , Xinchao Wang

Safety alignment in Large Language Models (LLMs) inherently presents a multi-objective optimization conflict, often accompanied by an unintended degradation of general capabilities. Existing mitigation strategies typically rely on global…

Machine Learning · Computer Science 2026-01-09 Wang Cai , Yilin Wen , Jinchang Hou , Du Su , Guoqiu Wang , Zhonghou Lv , Chenfu Bao , Yunfang Wu

Finetuning large language models (LLMs) enables user-specific customization but introduces critical safety risks: even a few harmful examples can compromise safety alignment. A common mitigation strategy is to update the model more strongly…

Machine Learning · Computer Science 2025-12-23 ShengYun Peng , Pin-Yu Chen , Jianfeng Chi , Seongmin Lee , Duen Horng Chau

Large language models (LLMs) are increasingly deployed in high-stakes domains, yet a unified treatment of their overlapping safety challenges remains lacking. We present SafeLM, a framework that jointly addresses four pillars of LLM safety:…

Cryptography and Security · Computer Science 2026-04-21 Noor Islam S. Mohammad , Uluğ Bayazıt

The growing awareness of safety concerns in large language models (LLMs) has sparked considerable interest in the evaluation of safety. This study investigates an under-explored issue about the evaluation of LLMs, namely the substantial…

Computation and Language · Computer Science 2024-04-02 Yixu Wang , Yan Teng , Kexin Huang , Chengqi Lyu , Songyang Zhang , Wenwei Zhang , Xingjun Ma , Yu-Gang Jiang , Yu Qiao , Yingchun Wang

With rapid advancement and increasing accessibility of LLMs, fine-tuning aligned models has become a critical step for adapting them to real-world applications, which makes the safety of this fine-tuning process more important than ever.…

Cryptography and Security · Computer Science 2025-07-28 Hao Li , Lijun Li , Zhenghao Lu , Xianyi Wei , Rui Li , Jing Shao , Lei Sha

Fine-tuning enables large language models (LLMs) to adapt to specific domains, but often compromises their previously established safety alignment. To mitigate the degradation of model safety during fine-tuning, we introduce LookAhead…

Computation and Language · Computer Science 2025-12-22 Kangwei Liu , Mengru Wang , Yujie Luo , Lin Yuan , Mengshu Sun , Lei Liang , Zhiqiang Zhang , Jun Zhou , Bryan Hooi , Shumin Deng

LLMs(Large Language Models) nowadays have widespread adoption as a tool for solving issues across various domain/tasks. These models since are susceptible to produce harmful or toxic results, inference-time adversarial attacks, therefore…

Computation and Language · Computer Science 2024-10-01 Sachin Kumar

Safety post-training can improve the harmfulness and policy compliance of Large Language Models (LLMs), but it may also reduce general utility, a phenomenon often described as the \emph{alignment tax}. We study this trade-off through the…

Machine Learning · Computer Science 2026-05-13 Guanglong Sun , Siyuan Zhang , Liyuan Wang , Jun Zhu , Hang Su , Yi Zhong

Large Language Models (LLMs) are increasingly integrated into educational applications. However, they remain vulnerable to jailbreak and fine-tuning attacks, which can compromise safety alignment and lead to harmful outputs. Existing…

Computation and Language · Computer Science 2025-11-19 Xin Yi , Yue Li , Dongsheng Shi , Linlin Wang , Xiaoling Wang , Liang He