Modern deployments require LLMs to enforce safety policies at scale, yet many controls rely on inference-time interventions that add recurring compute cost and serving complexity. Activation steering is widely used, but it requires runtime hooks and scales cost with the number of generations; conditional variants improve selectivity by gating when steering is applied but still retain an inference-time control path. We ask whether selective refusal can be moved entirely offline: can a mechanistic understanding of category-specific refusal be distilled into a circuit-restricted weight update that deploys as a standard checkpoint? We propose C-Δθ Circuit Restricted Weight Arithmetic}, which (i) localizes refusal-causal computation as a sparse circuit using EAP-IG and (ii) computes a constrained weight update ΔθC supported only on that circuit (typically <5% of parameters). Applying ΔθC yields a drop-in edited checkpoint with no inference-time hooks, shifting cost from per request intervention to a one-time offline update. We evaluate category-targeted selectivity and capability retention on refusal and utility benchmarks.
Utilize the Flow before Stepping into the Same River Twice: Certainty Represented Knowledge Flow for Refusal-Aware Instruction Tuning
Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）；Peking University（北京大学）；Southwest Jiaotong University（西南交通大学）；Hangzhou Institute for Advanced Study, University of Chinese Academy of Sciences（中国科学院大学杭州高等研究院）
Trusted Uncertainty in Large Language Models: A Unified Framework for Confidence Calibration and Risk-Controlled Refusal
University of Liechtenstein（列支敦士登大学）；University of the Republic of San Marino（圣马尔科共和国大学）；University of Mauritius（毛里求斯大学）；International University of Monaco（摩纳哥国际大学）；University of Andorra（安道尔大学）；University of Gibraltar（直布罗陀大学）；University of the Faroe Islands（法罗群岛大学）；Ilisimatusarfik (University of Greenland)（格陵兰岛研究所（格陵兰大学））；University of Prizren（普里兹伦大学）
HiddenGuard: Fine-Grained Safe Generation with Specialized Representation Router
Institute of Computing Technology, CAS（中国科学院计算技术研究所）；University of Chinese Academy of Sciences（中国科学院大学）；University of California, Merced（加州大学默塞德分校）；HKUST（香港科技大学）
HiddenGuard 提出基于隐藏状态的 Prism 路由器，在令牌级别实时检测并编辑有害内容，实现细粒度安全生成，在保持信息量的同时 F1 超过 90%。
Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）；University of Chinese Academy of Sciences（中国科学院大学）；The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））
Modular Arithmetic: Language Models Solve Math Digit by Digit
German Research Center for AI (DFKI)（德国人工智能研究中心（DFKI））；Saarland University（萨尔兰大学）；Center for European Research in Trusted AI (CERTAIN)（可信人工智能欧洲研究中心（CERTAIN））
Interpreting and Improving Large Language Models in Arithmetic Calculation
University of Science and Technology of China（中国科学技术大学）；Alibaba Cloud（阿里云）；Hong Kong Baptist University（香港浸会大学）；Institute of Artificial Intelligence, Hefei Comprehensive National Science Center（合肥综合性国家科学中心人工智能研究院）
