安全对部署大型语言模型（LLMs）在现实应用中至关重要，但大多数现有方法依赖于大规模人工标注数据集和静态红队基准测试，这些方法成本高、难以扩展且难以适应模型行为的变化。此外，过于保守的安全机制可能会通过拒绝敏感但合法的查询来降低模型的实用性。我们引入了Self-MOA（Self Multi-Objective Alignment），一种利用自动评估器模型进行弱监督对齐的小型语言模型的全自动框架。Self-MOA operates as a closed loop that dynamically generates model-specific red team prompts, constructs preference data from model-generated responses, and aligns models via multi-objective preference optimization to jointly optimize for safety and helpfulness. Across multiple small language models and safety benchmarks, Self-MOA achieves a 12.41% improvement in safety while preserving helpfulness, using as little as 11 times less training data than human-supervised alignment baselines. These results demonstrate that adaptive, automated alignment can reduce the dependence on static, human-curated safety pipelines in resource-constrained settings.
英文摘要：
Safety alignment is critical for deploying large language models (LLMs) in real-world applications, yet most existing approaches rely on large human-annotated datasets and static red-teaming benchmarks that are costly, difficult to scale, and slow to adapt to evolving model behaviors. Moreover, overly conservative safety mechanisms can reduce model usefulness by rejecting sensitive but legitimate queries. We introduce Self-MOA (Self Multi-Objective Alignment), a fully automated framework for aligning small language models using weak supervision from automated evaluator models. Self-MOA operates as a closed loop that dynamically generates model-specific red team prompts, constructs preference data from model-generated responses, and aligns models via multi-objective preference optimization to jointly optimize for safety and helpfulness. Across multiple small language models and safety benchmarks, Self-MOA achieves a 12.41\% improvement in safety while preserving helpfulness, using as little as 11 times less training data than human-supervised alignment baselines. These results demonstrate that adaptive, automated alignment can reduce the dependence on static, human-curated safety pipelines in resource-constrained settings.
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）
StepGuard: Learning Step-Level Guardrails with Scalable Supervision and Safety-Utility Balancing
Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）；Beihang University（北京航空航天大学）；Fudan University（复旦大学）；Renmin University of China（中国人民大学）；KAUST（阿卜杜拉国王科技大学）
SafeMed-R1: Clinician-Audited Safety and Ethics Alignment for Medical Large Language Models
Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）；Joint Laboratory of Biomedical Artificial Intelligence（生物医学人工智能联合实验室）；Shanghai Institute of Infectious Disease and Biosecurity（上海传染病与生物安全研究院）；Shanghai Health Development Research Center (Shanghai Medical Information Center)（上海健康发展战略研究中心（上海医疗信息中心））；University of Washington（华盛顿大学）；Department of Eye and Vision Sciences, University of Liverpool（利物浦大学眼科与视觉科学系）；Liverpool Centre for Cardiovascular Science, University of Liverpool（利物浦大学心血管科学中心）；School of Computer Science and Technology, Tongji University（同济大学计算机科学与技术学院）
Advancing LLM Safe Alignment with Safety Representation Ranking
State Key Lab of General Artificial Intelligence, School of Intelligence Science and Technology, Peking University（通用人工智能国家重点实验室，智能科学与技术学院，北京大学）；School of Mathematical Sciences, Peking University（数学科学学院，北京大学）；Institute for Artificial Intelligence, Peking University（人工智能研究院，北京大学）
Institute of Trustworthy Embodied AI（可信具身人工智能研究院）；Fudan University（复旦大学）；Shanghai Innovation Institute（上海创新研究院）；Ant Group（蚂蚁集团）；Zhejiang University（浙江大学）；City University of Hong Kong（香港城市大学）
From Prompts to Protection: Large Language Model-Enabled In-Context Learning for Smart Public Safety UAV
Real-Time and Embedded Computing Systems Research Centre (CISTER)（实时嵌入式计算系统研究中心）；Carnegie Mellon University（卡内基梅隆大学）；Pontificia Universidad Católica de Chile（天主教大学）；School of Computer Science, McGill University（麦吉尔大学计算机科学学院）；Instituto de Telecomunicações, Faculdade de Engenharia, Universidade do Porto（电信研究所，工程学院，葡萄牙里斯本大学）；University of Houston（休斯顿大学）
When Reasoning Is Not Enough: Proactive Inquiry for Cross-Modal Safety in Multimodal Large Language Models
Institute of Artificial Intelligence (TeleAI), China Telecom（中国电信人工智能研究院（TeleAI））；Peking University（北京大学）；School of Artificial Intelligence, OPtics and ElectroNics (iOPEN), Northwestern Polytechnical University（西北工业大学人工智能与光电子学院（iOPEN））
Self-Guided Adaptive Safety Alignment: Synthesizing and Internalizing Guidelines in Reasoning Models
Beijing Key Lab of Traffic Data Analysis and Mining（北京交通大数据分析与挖掘重点实验室）；Beijing Jiaotong University（北京交通大学）；School of Information Technology and Management（信息科技与管理学院）；University of International Business and Economics（国际经济贸易大学）
Can Vision-Language Models Stay Helpful When Facing Implicit Risks? Intent-Privilege OPSD for Efficient Safety-Helpfulness Alignment
Southern University of Science and Technology（南方科技大学）；Sun Yat-sen University（中山大学）；Guangdong Laboratory of Artificial Intelligence and Digital Economy (SZ)（广东省人工智能与数字经济实验室（深圳））；University of Electronic Science and Technology of China（电子科技大学）；Hangzhou Dianzi University（杭州电子科技大学）；Shenzhen MSU-BIT University（深圳北理莫斯科大学）；GigaAI（极佳科技）；The Chinese University of Hong Kong (Shenzhen)（香港中文大学（深圳））
