Hallucination in large language models (LLMs) has been widely studied in recent years, with progress in both detection and mitigation aimed at improving truthfulness. Yet, a critical side effect remains largely overlooked: enhancing truthfulness can negatively impact safety alignment. In this paper, we investigate this trade-off and show that increasing factual accuracy often comes at the cost of weakened refusal behavior. Our analysis reveals that this arises from overlapping components in the model that simultaneously encode hallucination and refusal information, leading alignment methods to suppress factual knowledge unintentionally. We further examine how fine-tuning on benign datasets, even when curated for safety, can degrade alignment for the same reason. To address this, we propose a method that disentangles refusal-related features from hallucination features using sparse autoencoders, and preserves refusal behavior during fine-tuning through subspace orthogonalization. This approach prevents hallucinations from increasing while maintaining safety alignment.We evaluate our method on commonsense reasoning tasks and harmful benchmarks (AdvBench and StrongReject). Results demonstrate that our approach preserves refusal behavior and task utility, mitigating the trade-off between truthfulness and safety.
Wisdom is Knowing What not to Say: Hallucination-Free LLMs Unlearning via Attention Shifting
Faculty of Information Technology, Monash University（墨尔本大学信息科技学院）；Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Shandong Computer Science Center, Qilu University of Technology (Shandong Academy of Sciences)（教育部计算能力网络与信息安全重点实验室，山东计算机科学中心，齐鲁大学（山东科学院））；Shandong Provincial Key Laboratory of Computing Power Internet and Service Computing, Shandong Fundamental Research Center for Computer Science（山东省计算能力互联网与服务计算重点实验室，山东省计算机科学基础研究中心）；School of Computer Science and Technology, Anhui University（安徽大学计算机科学与技术学院）
Balancing Efficiency and Efficacy: Training-Free Attention-Guided Switching Between Explicit and Latent Thoughts for MLLMs
Harbin Institute of Technology, Shenzhen（哈尔滨工业大学（深圳））；Tsinghua Shenzhen International Graduate School, Tsinghua University（清华大学深圳国际研究生院）；Peng Cheng Laboratory（鹏城实验室）；Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences（中国科学院深圳先进技术研究院）
National Engineering Research Center of Speech and Language Information Processing（语音与语言信息处理国家级工程研究中心）；University of Science and Technology of China（中国科学技术大学）；iFLYTEK Research（iFLYTEK研究院）；University of California, Los Angeles（美国加州大学洛杉矶分校）
Medication counseling with large language models: balancing flexibility and rigidity
Department of Engineering and Information Technology Åbo Akademi University（工程与信息科技系阿博阿卡迪米大学）；Experience Lab Åbo Akademi University（经验实验室阿博阿卡迪米大学）；Department of Natural and Health Sciences Åbo Akademi University（自然与健康科学系阿博阿卡迪米大学）；Department of Caring and Ethics University of Stavanger（护理与伦理系斯塔万格大学）
Enhancing LLM Reliability via Explicit Knowledge Boundary Modeling
X-LANCE Lab, School of Computer Science MoE Key Lab of Artificial Intelligence, SJTU AI Institute Shanghai Jiao Tong University, Shanghai, China（X-LANCE实验室，计算机科学学院，人工智能关键实验室，上海交通大学人工智能研究院，上海，中国）；Jiangsu Key Lab of Language Computing, Suzhou, China（语言计算重点实验室，苏州，中国）；AISpeech Co., Ltd., Suzhou, China（AISpeech公司，苏州，中国）；Center for Artificial Intelligence Research (CAiRE) Hong Kong University of Science and Technology, Hong Kong, China（人工智能研究中心（CAiRE）香港科技大学，香港，中国）；Suzhou Laboratory, Suzhou, China（苏州实验室，苏州，中国）
