English
Related papers

Related papers: The Geometry of Alignment Collapse: When Fine-Tuni…

200 papers

Aligned LLMs are secure, capable of recognizing and refusing to answer malicious questions. However, the role of internal parameters in maintaining such security is not well understood yet, further these models can be vulnerable to security…

Cryptography and Security · Computer Science 2025-04-08 Shen Li , Liuyi Yao , Lan Zhang , Yaliang Li

Current Large Language Models (LLMs), even those tuned for safety and alignment, are susceptible to jailbreaking. Some have found that just further fine-tuning an aligned model with benign data (i.e., data without harmful content)…

Machine Learning · Computer Science 2024-08-21 Luxi He , Mengzhou Xia , Peter Henderson

Large language models require continuous adaptation to new tasks while preserving safety alignment. However, fine-tuning on even benign data often compromises safety behaviors, including refusal of harmful requests, truthfulness, and…

Machine Learning · Computer Science 2026-04-21 Thong Bach , Dung Nguyen , Thao Minh Le , Truyen Tran

A guard model fine-tuned on entirely benign data can lose all safety alignment -- not through adversarial manipulation, but through standard domain specialization. We demonstrate this failure across three purpose-built safety classifiers --…

The deployment of large language models (LLMs) raises significant ethical and safety concerns. While LLM alignment techniques are adopted to improve model safety and trustworthiness, adversaries can exploit these techniques to undermine…

Cryptography and Security · Computer Science 2026-04-10 Rui Zhang , Hongwei Li , Yun Shen , Xinyue Shen , Wenbo Jiang , Guowen Xu , Yang Liu , Michael Backes , Yang Zhang

Fine-tuning is arguably the most straightforward way to tailor a pre-trained model (e.g., a foundation model) to downstream applications, but it also comes with the risk of losing valuable knowledge the model had learned in pre-training.…

Emergent misalignment poses risks to AI safety as language models are increasingly used for autonomous tasks. In this paper, we present a population of large language models (LLMs) fine-tuned on insecure datasets spanning 11 diverse…

Artificial Intelligence · Computer Science 2026-02-03 Abhishek Mishra , Mugilan Arulvanan , Reshma Ashok , Polina Petrova , Deepesh Suranjandass , Donnie Winkelmann

As large language models (LLMs) gain popularity, their vulnerability to adversarial attacks emerges as a primary concern. While fine-tuning models on domain-specific datasets is often employed to improve model performance, it can…

Computation and Language · Computer Science 2026-02-24 Punya Syon Pandey , Samuel Simko , Kellin Pelrine , Zhijing Jin

Fine-tuning lets practitioners repurpose aligned large language models (LLMs) for new domains, yet recent work reveals emergent misalignment (EMA): Even a small, domain-specific fine-tune can induce harmful behaviors far outside the target…

Machine Learning · Computer Science 2026-03-06 David Kaczér , Magnus Jørgenvåg , Clemens Vetter , Esha Afzal , Robin Haselhorst , Lucie Flek , Florian Mai

Explicit reasoning models are trained to produce intermediate reasoning traces before final answers, but downstream fine-tuning is often performed on ordinary instruction-response data that contains no such traces. We show that this…

Machine Learning · Computer Science 2026-05-21 Lukas Twist , Helen Yannakoudakis , Jie M. Zhang

Organizations are increasingly adopting and adapting Large Language Models (LLMs) hosted on public repositories such as HuggingFace. Although these adaptations often improve performance on specialized downstream tasks, recent evidence…

Artificial Intelligence · Computer Science 2025-11-04 Mina Taraghi , Yann Pequignot , Amin Nikanjam , Mohamed Amine Merzouk , Foutse Khomh

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

Harmful fine-tuning attack poses serious safety concerns for large language models' fine-tuning-as-a-service. While existing defenses have been proposed to mitigate the issue, their performances are still far away from satisfactory, and the…

Computation and Language · Computer Science 2025-03-18 Tiansheng Huang , Sihao Hu , Fatih Ilhan , Selim Furkan Tekin , Ling Liu

Fine-tuning text-to-image diffusion models is widely used for personalization and adaptation for new domains. In this paper, we identify a critical vulnerability of fine-tuning: safety alignment methods designed to filter harmful content…

Artificial Intelligence · Computer Science 2024-12-03 Sanghyun Kim , Moonseok Choi , Jinwoo Shin , Juho Lee

Fine-tuning is an essential and pervasive functionality for applying large language models (LLMs) to downstream tasks. However, it has the potential to substantially degrade safety alignment, e.g., by greatly increasing susceptibility to…

Machine Learning · Computer Science 2026-01-16 Jiawen Zhang , Yangfan Hu , Kejia Chen , Lipeng He , Jiachen Ma , Jian Lou , Dan Li , Jian Liu , Xiaohu Yang , Ruoxi Jia

Fine-tuning large language models (LLMs) for downstream tasks often leads to catastrophic forgetting, notably degrading the safety of originally aligned models. While some existing methods attempt to restore safety by incorporating…

Computation and Language · Computer Science 2025-08-29 Hua Farn , Hsuan Su , Shachi H Kumar , Saurav Sahay , Shang-Tse Chen , Hung-yi Lee

Weight-perturbation evolution strategies (ES) can fine-tune billion-parameter language models with surprisingly small populations (e.g., $N\!\approx\!30$), contradicting classical zeroth-order curse-of-dimensionality intuition. We also…

Machine Learning · Computer Science 2026-02-03 Qiyao Liang , Jinyeop Song , Yizhou Liu , Jeff Gore , Ila Fiete , Risto Miikkulainen , Xin Qiu

The safety alignment of large language models (LLMs) is becoming increasingly important with their democratization. In this paper, we study the safety degradation that comes with adapting LLMs to new tasks. We attribute this safety…

Computation and Language · Computer Science 2025-12-12 Lama Alssum , Hani Itani , Hasan Abed Al Kader Hammoud , Philip Torr , Adel Bibi , Bernard Ghanem

We discover the emergence of \textit{basins} in the loss landscape of large language models. As model scale increases, LLMs become progressively more resilient to random perturbations in the parameter space, giving rise to expansive…

Machine Learning · Computer Science 2026-03-17 Huanran Chen , Yinpeng Dong , Zeming Wei , Yao Huang , Yichi Zhang , Hang Su , Jun Zhu

Fine-tuning Large Language Models (LLMs) on some task-specific datasets has been a primary use of LLMs. However, it has been empirically observed that this approach to enhancing capability inevitably compromises safety, a phenomenon also…

Machine Learning · Statistics 2025-03-28 Pin-Yu Chen , Han Shen , Payel Das , Tianyi Chen