SafeSwitch: Steering Unsafe LLM Behavior via Internal Activation Signals
Siebel School of Computing(计算科学系);Data Science, University of Illinois Urbana-Champaign(数据科学,伊利诺伊大学厄巴纳-香槟分校);School of Business, Stevens Institute of Technology(商学院,史蒂文斯理工学院)
Large language models (LLMs) exhibit exceptional capabilities across various tasks but also pose risks by generating harmful content. Existing safety mechanisms, while improving model safety, often lead to overly cautious behavior and fail to fully leverage LLMs' internal cognitive processes. Inspired by humans' reflective thinking capability, we first show that LLMs can similarly perform internal assessments about safety in their internal states. Building on this insight, we propose SafeSwitch, a dynamic framework that regulates unsafe outputs by utilizing the prober-based internal state monitor that actively detects harmful intentions, and activates a safety head that leads to safer and more conservative responses only when necessary. SafeSwitch reduces harmful outputs by approximately 80% on harmful queries while maintaining strong utility, reaching a Pareto optimal among several methods. Our method is also advantageous over traditional methods in offering more informative, context-aware refusals, and achieves these benefits while only tuning less than 6% of the original parameters. SafeSwitch demonstrates large language models' capacity for self-awareness and reflection regarding safety, offering a promising approach to more nuanced and effective safety controls. Codes for this work are available at https://github.com/Hanpx20/SafeSwitch.
BarrierSteer: LLM Safety via Learning Barrier Steering
Department of Computer Science, National University of Singapore(新加坡国立大学计算机科学系);Singapore-MIT Alliance for Research and Technology Centre(新加坡-麻省理工联合研究中心);CSAIL, Massachusetts Institute of Technology(麻省理工学院计算机科学与人工智能实验室);Worcester Polytechnic Institute(沃斯堡理工学院)
Activation Steering Induces Emergent Misalignment: Steering Vector Construction and Comprehensive Evaluation
Nanyang Technological University(南洋理工大学);Sun Yat-sen University(中山大学);University of Science and Technology of China(中国科学技术大学);National University of Singapore(新加坡国立大学)
When Do Model Internals Help? Exploring the Role of Representation Engineering in LLM Safety
State Key Laboratory of Multimedia Information Processing, Peking University(北京大学多媒体信息处理国家重点实验室);School of Computer Science, Peking University(北京大学计算机科学学院)
Dialectics of Alignment: Harnessing Unsafe Knowledge for Dynamic Safety Routing
Chandar Research Lab(Chandar研究实验室);Mila – Quebec AI Institute(魁北克AI研究所);Université de Montréal(蒙特利尔大学);Microsoft Research(微软研究院);Polytechnique Montréal(蒙特利尔理工学院);Canada CIFAR AI Chair(加拿大CIFAR人工智能主席)
Selective Safety Steering via Value-Filtered Decoding
Department of Electrical and Computer Engineering, Technion IIT(技术学院电气与计算机工程系);Department of Statistics, University of Oxford(牛津大学统计系);OATML, Department of Computer Science, University of Oxford(牛津大学计算机科学系);Department of Computer Science, Technion IIT(技术学院计算机科学系)
Internal Activation Revision: Safeguarding Vision Language Models Without Parameter Update
Mohamed bin Zayed University of Artificial Intelligence(穆罕默德·本·扎耶德人工智能大学);Fraunhofer FOKUS(弗劳恩霍夫协会FOKUS研究所);The University of Tokyo(东京大学);University of Alberta(阿尔伯塔大学)
Probing the Robustness of Large Language Models Safety to Latent Perturbations
Tsinghua Shenzhen International Graduate School, Tsinghua University(清华大学深圳国际 Graduate School, 清华大学);Shanghai Artificial Intelligence Laboratory(上海人工智能实验室);Fudan University(复旦大学);The University of Hong Kong(香港大学)