_version_ 1866908468341899264
author Lu, Haoran
Fang, Luyang
Zhang, Ruidong
Li, Xinliang
Cai, Jiazhang
Cheng, Huimin
Tang, Lin
Liu, Ziyu
Sun, Zeliang
Wang, Tao
Zhang, Yingchuan
Zidan, Arif Hassan
Xu, Jinwen
Yu, Jincheng
Yu, Meizhi
Jiang, Hanqi
Gong, Xilin
Luo, Weidi
Sun, Bolun
Chen, Yongkai
Ma, Terry
Wu, Shushan
Zhou, Yifan
Chen, Junhao
Xiang, Haotian
Zhang, Jing
Jahin, Afrar
Ruan, Wei
Deng, Ke
Pan, Yi
Wang, Peilong
Li, Jiahui
Liu, Zhengliang
Zhang, Lu
Zhao, Lin
Liu, Wei
Zhu, Dajiang
Xing, Xin
Dou, Fei
Zhang, Wei
Huang, Chao
Liu, Rongjie
Zhang, Mengrui
Liu, Yiwen
Sun, Xiaoxiao
Lu, Qin
Xiang, Zhen
Zhong, Wenxuan
Liu, Tianming
Ma, Ping
author_facet Lu, Haoran
Fang, Luyang
Zhang, Ruidong
Li, Xinliang
Cai, Jiazhang
Cheng, Huimin
Tang, Lin
Liu, Ziyu
Sun, Zeliang
Wang, Tao
Zhang, Yingchuan
Zidan, Arif Hassan
Xu, Jinwen
Yu, Jincheng
Yu, Meizhi
Jiang, Hanqi
Gong, Xilin
Luo, Weidi
Sun, Bolun
Chen, Yongkai
Ma, Terry
Wu, Shushan
Zhou, Yifan
Chen, Junhao
Xiang, Haotian
Zhang, Jing
Jahin, Afrar
Ruan, Wei
Deng, Ke
Pan, Yi
Wang, Peilong
Li, Jiahui
Liu, Zhengliang
Zhang, Lu
Zhao, Lin
Liu, Wei
Zhu, Dajiang
Xing, Xin
Dou, Fei
Zhang, Wei
Huang, Chao
Liu, Rongjie
Zhang, Mengrui
Liu, Yiwen
Sun, Xiaoxiao
Lu, Qin
Xiang, Zhen
Zhong, Wenxuan
Liu, Tianming
Ma, Ping
contents Due to the remarkable capabilities and growing impact of large language models (LLMs), they have been deeply integrated into many aspects of society. Thus, ensuring their alignment with human values and intentions has emerged as a critical challenge. This survey provides a comprehensive overview of practical alignment techniques, training protocols, and empirical findings in LLM alignment. We analyze the development of alignment methods across diverse paradigms, characterizing the fundamental trade-offs between core alignment objectives. Our analysis shows that while supervised fine-tuning enables basic instruction-following, preference-based methods offer more flexibility for aligning with nuanced human intent. We discuss state-of-the-art techniques, including Direct Preference Optimization (DPO), Constitutional AI, brain-inspired methods, and alignment uncertainty quantification (AUQ), highlighting their approaches to balancing quality and efficiency. We review existing evaluation frameworks and benchmarking datasets, emphasizing limitations such as reward misspecification, distributional robustness, and scalable oversight. We summarize strategies adopted by leading AI labs to illustrate the current state of practice. We conclude by outlining open problems in oversight, value pluralism, robustness, and continuous alignment. This survey aims to inform both researchers and practitioners navigating the evolving landscape of LLM alignment.
format Preprint
id arxiv_https___arxiv_org_abs_2507_19672
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Alignment and Safety in Large Language Models: Safety Mechanisms, Training Paradigms, and Emerging Challenges
Lu, Haoran
Fang, Luyang
Zhang, Ruidong
Li, Xinliang
Cai, Jiazhang
Cheng, Huimin
Tang, Lin
Liu, Ziyu
Sun, Zeliang
Wang, Tao
Zhang, Yingchuan
Zidan, Arif Hassan
Xu, Jinwen
Yu, Jincheng
Yu, Meizhi
Jiang, Hanqi
Gong, Xilin
Luo, Weidi
Sun, Bolun
Chen, Yongkai
Ma, Terry
Wu, Shushan
Zhou, Yifan
Chen, Junhao
Xiang, Haotian
Zhang, Jing
Jahin, Afrar
Ruan, Wei
Deng, Ke
Pan, Yi
Wang, Peilong
Li, Jiahui
Liu, Zhengliang
Zhang, Lu
Zhao, Lin
Liu, Wei
Zhu, Dajiang
Xing, Xin
Dou, Fei
Zhang, Wei
Huang, Chao
Liu, Rongjie
Zhang, Mengrui
Liu, Yiwen
Sun, Xiaoxiao
Lu, Qin
Xiang, Zhen
Zhong, Wenxuan
Liu, Tianming
Ma, Ping
Artificial Intelligence
Machine Learning
Due to the remarkable capabilities and growing impact of large language models (LLMs), they have been deeply integrated into many aspects of society. Thus, ensuring their alignment with human values and intentions has emerged as a critical challenge. This survey provides a comprehensive overview of practical alignment techniques, training protocols, and empirical findings in LLM alignment. We analyze the development of alignment methods across diverse paradigms, characterizing the fundamental trade-offs between core alignment objectives. Our analysis shows that while supervised fine-tuning enables basic instruction-following, preference-based methods offer more flexibility for aligning with nuanced human intent. We discuss state-of-the-art techniques, including Direct Preference Optimization (DPO), Constitutional AI, brain-inspired methods, and alignment uncertainty quantification (AUQ), highlighting their approaches to balancing quality and efficiency. We review existing evaluation frameworks and benchmarking datasets, emphasizing limitations such as reward misspecification, distributional robustness, and scalable oversight. We summarize strategies adopted by leading AI labs to illustrate the current state of practice. We conclude by outlining open problems in oversight, value pluralism, robustness, and continuous alignment. This survey aims to inform both researchers and practitioners navigating the evolving landscape of LLM alignment.
title Alignment and Safety in Large Language Models: Safety Mechanisms, Training Paradigms, and Emerging Challenges
topic Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2507.19672