中文
相关论文

相关论文: Eliciting and Analyzing Emergent Misalignment in S…

200 篇论文

Large language models (LLMs) offer transformative potential for clinical decision support in spine surgery but pose significant risks through hallucinations, which are factually inconsistent or contextually misaligned outputs that may…

机器学习 · 计算机科学 2025-11-21 Dong Chen , Yanzhe Wei , Zonglin He , Guan-Ming Kuang , Canhua Ye , Meiru An , Huili Peng , Yong Hu , Huiren Tao , Kenneth MC Cheung

We stress-tested 16 leading models from multiple developers in hypothetical corporate environments to identify potentially risky agentic behaviors before they cause real harm. In the scenarios, we allowed models to autonomously send emails…

密码学与安全 · 计算机科学 2025-10-17 Aengus Lynch , Benjamin Wright , Caleb Larson , Stuart J. Ritchie , Soren Mindermann , Evan Hubinger , Ethan Perez , Kevin Troy

Large language models (LLMs) have made remarkable strides in complex reasoning tasks, but their safety and robustness in reasoning processes remain underexplored. Existing attacks on LLM reasoning are constrained by specific settings or…

人工智能 · 计算机科学 2025-06-17 Jingyu Peng , Maolin Wang , Xiangyu Zhao , Kai Zhang , Wanyu Wang , Pengyue Jia , Qidong Liu , Ruocheng Guo , Qi Liu

Language model agents excel in long-session planning and reasoning, but existing benchmarks primarily focus on goal-oriented tasks with explicit objectives, neglecting creative adaptation in unfamiliar environments. To address this, we…

As large language models (LLMs) rapidly displace traditional expertise, their capacity to correct misinformation has become a core concern. We investigate the idea that prompt framing systematically modulates misinformation correction -…

人机交互 · 计算机科学 2025-12-01 Sekoul Krastev , Hilary Sweatman , Anni Sternisko , Steve Rathje

While brain-aligned large language models (LLMs) have garnered attention for their potential as cognitive models and for potential for enhanced safety and trustworthiness in AI, the role of this brain alignment for linguistic competence…

计算与语言 · 计算机科学 2026-03-25 Gabriele Merlin , Mariya Toneva

Despite the remarkable advancements and widespread applications of deep neural networks, their ability to perform reasoning tasks remains limited, particularly in domains requiring structured, abstract thought. In this paper, we investigate…

计算与语言 · 计算机科学 2025-09-16 Satyam Goyal , Soham Dan

Current benchmarks for Large Language Models (LLMs) primarily focus on performance metrics, often failing to capture the nuanced behavioral characteristics that differentiate them. This paper introduces a novel ``Behavioral Fingerprinting''…

计算与语言 · 计算机科学 2025-09-08 Zehua Pei , Hui-Ling Zhen , Ying Zhang , Zhiyuan Yang , Xing Li , Xianzhi Yu , Mingxuan Yuan , Bei Yu

The safety and alignment of Large Language Models (LLMs) are critical for their responsible deployment. Current evaluation methods predominantly focus on identifying and preventing overtly harmful outputs. However, they often fail to…

Large language models (LLMs) have demonstrated remarkable capabilities across a wide range of natural language processing tasks. However, their tendency to exhibit sycophantic behavior - excessively agreeing with or flattering users - poses…

计算与语言 · 计算机科学 2025-01-30 Lars Malmqvist

Recent studies on the safety alignment of large language models (LLMs) have revealed that existing approaches often operate superficially, leaving models vulnerable to various adversarial attacks. Despite their significance, these studies…

密码学与安全 · 计算机科学 2025-06-02 Jianwei Li , Jung-Eun Kim

Large language models (LLMs) have been widely adopted as the core of agent frameworks in various scenarios, such as social simulations and AI companions. However, the extent to which they can replicate human-like motivations remains an…

计算与语言 · 计算机科学 2025-06-17 Xixian Yong , Jianxun Lian , Xiaoyuan Yi , Xiao Zhou , Xing Xie

Large Language Models(LLMs) have been successful in numerous fields. Alignment has usually been applied to prevent them from harmful purposes. However, aligned LLMs remain vulnerable to jailbreak attacks that deliberately mislead them into…

密码学与安全 · 计算机科学 2026-02-17 Shang Liu , Hanyu Pei , Zeyan Liu

With the development of large language models (LLMs) like ChatGPT, both their vast applications and potential vulnerabilities have come to the forefront. While developers have integrated multiple safety mechanisms to mitigate their misuse,…

计算与语言 · 计算机科学 2024-07-23 Xiao Liu , Liangzhi Li , Tong Xiang , Fuying Ye , Lu Wei , Wangyue Li , Noa Garcia

Large language models (LLMs) undergo alignment training to avoid harmful behaviors, yet the resulting safeguards remain brittle: jailbreaks routinely bypass them, and fine-tuning on narrow domains can induce ``emergent misalignment'' that…

Recent work has shown that fine-tuning large language models (LLMs) on code with security vulnerabilities can result in misaligned and unsafe behaviors across broad domains. These results prompted concerns about the emergence of harmful…

机器学习 · 计算机科学 2025-07-08 Jeremiah Giordani

Understanding how language models generalize behaviors from their training to a broader deployment distribution is an important problem in AI safety. Betley et al. discovered that fine-tuning GPT-4o on intentionally insecure code causes…

Large Language Models (LLMs) are expected to provide helpful and harmless responses, yet they often exhibit sycophancy--conforming to user beliefs regardless of factual accuracy or ethical soundness. Prior research on sycophancy has…

计算与语言 · 计算机科学 2026-03-02 Jiseung Hong , Grace Byun , Seungone Kim , Kai Shu , Jinho D. Choi

Large language models (LLMs) are increasingly deployed in a wide range of applications, yet remain vulnerable to adversarial jailbreak attacks that circumvent their safety guardrails. Existing evaluation frameworks typically report binary…

密码学与安全 · 计算机科学 2026-05-14 Zvi Topol

Large language models demonstrate strong reasoning capabilities through chain-of-thought prompting, but whether this reasoning quality transfers across languages remains underexplored. We introduce a human-validated framework to evaluate…

计算与语言 · 计算机科学 2026-03-31 Anaelia Ovalle , Candace Ross , Sebastian Ruder , Adina Williams , Karen Ullrich , Mark Ibrahim , Levent Sagun