在大语言模型（Large Language Models, LLM）的整个开发生命周期中，安全性始终是核心关切，研究人员与产业从业者愈发重视保障LLM的行为安全，使其与人类偏好及伦理标准对齐。基于海量多语言语料训练的LLM在不同语言与领域中展现出强大的泛化能力，但当前的安全对齐实践主要聚焦单语言场景，其在复杂多语言语境下，尤其是复杂混合语言格式中的有效性仍未得到充分探究。\n本研究提出Multilingual Blending（多语言混合）方案，这是一种混合语言的查询-响应机制，用于评估GPT-4o、GPT-3.5、Llama3等各类前沿LLM在复杂多语言条件下的安全对齐表现。我们进一步探究了语言可用性、形态学、语系等语言特征对Multilingual Blending突破LLM安全防护效果的影响。\n实验结果表明，无需精心设计提示模板，Multilingual Blending就能显著放大恶意查询的危害，导致LLM安全对齐的绕过率大幅上升：GPT-3.5的绕过率达67.23%，GPT-4o达40.34%，远高于单语言基线的水平。此外，Multilingual Blending的效果因语言内在属性的不同存在显著差异，不同形态、不同语系的语言更易规避安全对齐。这些发现凸显了在复杂多语言语境下评估LLM并制定相应安全对齐策略的必要性，以匹配其卓越的跨语言泛化能力。
英文摘要：
As safety remains a crucial concern throughout the development lifecycle of Large Language Models (LLMs), researchers and industrial practitioners have increasingly focused on safeguarding and aligning LLM behaviors with human preferences and ethical standards. LLMs, trained on extensive multilingual corpora, exhibit powerful generalization abilities across diverse languages and domains. However, current safety alignment practices predominantly focus on single-language scenarios, which leaves their effectiveness in complex multilingual contexts, especially for those complex mixed-language formats, largely unexplored. In this study, we introduce Multilingual Blending, a mixed-language query-response scheme designed to evaluate the safety alignment of various state-of-the-art LLMs (e.g., GPT-4o, GPT-3.5, Llama3) under sophisticated, multilingual conditions. We further investigate language patterns such as language availability, morphology, and language family that could impact the effectiveness of Multilingual Blending in compromising the safeguards of LLMs. Our experimental results show that, without meticulously crafted prompt templates, Multilingual Blending significantly amplifies the detriment of malicious queries, leading to dramatically increased bypass rates in LLM safety alignment (67.23% on GPT-3.5 and 40.34% on GPT-4o), far exceeding those of single-language baselines. Moreover, the performance of Multilingual Blending varies notably based on intrinsic linguistic properties, with languages of different morphology and from diverse families being more prone to evading safety alignments. These findings underscore the necessity of evaluating LLMs and developing corresponding safety alignment strategies in a complex, multilingual context to align with their superior cross-language generalization capabilities.
Beijing Academy of Artificial Intelligence（北京人工智能研究院）；National University of Singapore（新加坡国立大学）；Institute for Artificial Intelligence, Peking University（北京大学人工智能研究院）
MPO: Multilingual Safety Alignment via Reward Gap Optimization
Harbin Institute of Technology（哈尔滨工业大学）；Singapore Management University（新加坡管理学院）；Monash University（墨尔本大学）；Singapore University of Technology and Design（新加坡科技设计大学）；National University of Singapore（新加坡国立大学）
Qorgau: Evaluating LLM Safety in Kazakh-Russian Bilingual Contexts
Mohamed bin Zayed University of Artificial Intelligence（莫扎伊德大学人工智能学院）；Monash University（莫纳什大学）；University of Copenhagen（哥本哈根大学）；The University of Melbourne（墨尔本大学）；LibrAI（LibrAI公司）；Institute of Foundation Models（基础模型研究所）
City University of Hong Kong（香港城市大学）；Rutgers University（罗格斯大学）；University of Science and Technology of China（中国科学技术大学）；Universiteit van Amsterdam（阿姆斯特丹大学）；Baidu Inc（百度公司）
Exploring Adversarial Robustness and Safety Alignment in Multilingual Multi-Modal Large Language Models
Mohamed Bin Zayed University of AI, UAE（穆罕默德·本·扎耶德人工智能大学，阿联酋）；Khalifa University, UAE（卡比拉大学，阿联酋）；Australian National University, Australia（澳大利亚国立大学，澳大利亚）
Multilingual Safety Signals Are Multi-Layered: Filtering Safety-Degrading Data for Safer LLMs
Beihang University（北京航空航天大学）；Zhongguancun Laboratory（中关村实验室）；Hangzhou Innovation Institute, Beihang University（北京航空航天大学杭州创新研究院）；Qingdao Research Institute, Beihang University（北京航空航天大学青岛研究院）
SurakshaEval: An Indic Safety Benchmark for Multilingual LLMs
Inception42；Mohamed bin Zayed University of Artificial Intelligence（穆罕默德·本·扎耶德人工智能大学）；University of Michigan（密歇根大学）；Institute of Computer Science Artificial Intelligence & Technology（计算机科学人工智能与技术研究院）；International Institute of Information Technology Hyderabad（海得拉巴国际信息技术学院）；Ministry of Electronics and IT India（印度电子与信息技术部）；Indian Institute of Technology Kharagpur（印度理工学院卡拉格普尔分校）
XL-SafetyBench: A Country-Grounded Cross-Cultural Benchmark for LLM Safety and Cultural Sensitivity
AIM Intelligence（AIM智能研究院）；Microsoft（微软公司）；Korea AISI（韩国人工智能研究所）；KT Corporation（KT公司）；BMW Group（宝马集团）；Coinbase（Coinbase公司）；Technical University of Munich（慕尼黑技术大学）；Ankara University（安卡拉大学）；Cyril Amarchand Mangaldas（Cyril Amarchand Mangaldas法律事务所）；Seoul National University（首尔国立大学）
LLM Safety Alignment in Low-Resource Languages: A Systematic Literature Review
African Institute for Mathematical Sciences (AIMS)（非洲数学科学研究所）；Bayero University Kano（卡诺贝罗大学）；Brown University（布朗大学）；Centrum Wiskunde & Informatica（数学与计算机科学中心）；University of Pretoria（比勒陀利亚大学）；University of Hamburg（汉堡大学）；Imperial College London（伦敦帝国学院）
Towards Understanding the Fragility of Multilingual LLMs against Fine-Tuning Attacks
Meta；University of Pisa（比萨大学）；University of Modena and Reggio Emilia（摩德纳与雷焦艾米利亚大学）；Brown University（布朗大学）；University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
Multilingual Collaborative Defense for Large Language Models
Key Laboratory of Big Data & Artificial Intelligence in Transportation (Beijing Jiaotong University), Ministry of Education（大数据与人工智能交通运输 key laboratory（北京交通大学））；School of Computer Science and Technology, Beijing Jiaotong University（计算机科学与技术学院（北京交通大学））；University of Montreal（蒙特利尔大学）
