Many safety post-training methods for large language models (LLMs) are designed to modify the model's behaviour from producing unsafe answers to issuing refusals. However, such distribution shifts are often brittle and degrade performance on desirable tasks. To address these pitfalls, we propose augmenting the model's vocabulary with a special red flag token, and training the model to insert this token whenever harmful content is generated or imminent. This approach enables the model to explicitly learn the concept of harmfulness in its representations, with minimal impact on utility due to the marginal change in the generated distribution of natural language. Moreover, because the token is embedded in the model's vocabulary, we can naturally leverage the LLMs' generalization capabilities, such as in-context learning (ICL) and out-of-distribution generalization to languages that are not formally supported (e.g., Japanese for Llama3). In particular, we demonstrate that through ICL alone, the model can learn to initiate reflective reasoning upon generating the red flag token at inference, which steers the response away from harmful continuations or enables self-correction when the flag is raised falsely. This approach is orthogonal and complementary to existing safety technique (such as safety classifiers or standard safety training) and easier to evaluate in comparison to natural language refusals, as it does not require a human or automated judge to assess the harmlessness of the answers.
Gradual Vigilance and Interval Communication: Enhancing Value Alignment in Multi-Agent Debates
National Engineering Research Center of Educational Big Data, Central China Normal University（华中师范大学教育大数据国家工程研究中心）；Faculty of Artificial Intelligence in Education, Central China Normal University（华中师范大学人工智能教育学部）
Adversarial Preference Learning for Robust LLM Alignment
Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）；Northeastern University（东北大学）；University of Science and Technology of China（中国科学技术大学）；MemTensor (Shanghai) Technology Co., Ltd（MemTensor（上海）技术有限公司）；University of Auckland（奥克兰大学）；China Telecom Corporation Limited Beijing Research Institute（中国电信股份有限公司北京研究院）
Fake News Detectors are Biased against Texts Generated by Large Language Models
Mohamed bin Zayed University of Artificial Intelligence（穆罕默德·本·扎耶德人工智能大学）；King Abdullah University of Science and Technology（阿卜杜拉国王科技大学）；Monash University（莫纳什大学）；CSIRO’s Data61（澳大利亚联邦科学与工业研究组织Data61部门）
ResidencyRL: Reinforcement Learning in Simulated Clinical Environments
Google DeepMind（谷歌DeepMind）；Google Research（谷歌研究院）；Houston Methodist Hospital（休斯顿卫理公会医院）；Trinity Health Group（三一健康集团）；Stanford Oncology Partners（斯坦福肿瘤学伙伴）；St. Luke Hospital（圣卢克医院）
本研究提出 ResidencyRL，通过多轮强化学习训练临床 AI 智能体，在模拟临床环境中提升诊断准确性、降低漏报率，且能力可迁移至多个医学基准测试，为临床 AI 发展提供了新路径。
