中文
相关论文

相关论文: STAR: Detecting Inference-time Backdoors in LLM Re…

200 篇论文

With the growing popularity of LLM agents and RAG, it has become increasingly important to retrieve documents that are essential for solving a task, even when their connection to the task is indirect or implicit. Addressing this problem…

信息检索 · 计算机科学 2025-10-14 Junwei Lan , Jianlyu Chen , Zheng Liu , Chaofan Li , Siqi Bao , Defu Lian

Reverse engineering of protocol message formats is critical for many security applications. Mainstream techniques use dynamic analysis and inherit its low-coverage problem -- the inferred message formats only reflect the features of their…

密码学与安全 · 计算机科学 2024-07-02 Qingkai Shi , Xiangzhe Xu , Xiangyu Zhang

Can large language models introspect, that is, accurately detect perturbations to their own internal states? We systematically investigate this question using activation steering in Meta-Llama-3.1-8B-Instruct. First, we show that the binary…

人工智能 · 计算机科学 2026-03-03 Ely Hahami , Ishaan Sinha , Lavik Jain , Josh Kaplan , Jon Hahami

Large Language Models (LLMs) are widely deployed in diverse real-world settings, yet remain vulnerable to jailbreaking, where prompt-based attacks bypass safety filters. We present THREAT (Targeted Harmful generation via Reframing and…

密码学与安全 · 计算机科学 2026-05-22 Shahnewaz Karim Sakib , Swati Kar , Anindya Bijoy Das

Large Language Models (LLMs), despite advanced general capabilities, still suffer from numerous safety risks, especially jailbreak attacks that bypass safety protocols. Understanding these vulnerabilities through black-box jailbreak…

密码学与安全 · 计算机科学 2025-05-29 Yao Huang , Yitong Sun , Shouwei Ruan , Yichi Zhang , Yinpeng Dong , Xingxing Wei

Large language models (LLMs) increasingly solve difficult problems by producing "reasoning traces" before emitting a final response. However, it remains unclear how accuracy and decision commitment evolve along a reasoning trajectory, and…

机器学习 · 计算机科学 2026-02-02 Marthe Ballon , Brecht Verbeken , Vincent Ginis , Andres Algaba

Large language models frequently commit unrecoverable reasoning errors mid-generation: once a wrong step is taken, subsequent tokens compound the mistake rather than correct it. We introduce $\textbf{Latent Phase-Shift Rollback}$ (LPSR): at…

机器学习 · 计算机科学 2026-04-21 Manan Gupta , Dhruv Kumar

In reasoning chains generated by large language models (LLMs), initial errors often propagate and undermine the reliability of the final conclusion. Current LLM-based error detection methods often fail to detect propagated errors because…

机器学习 · 计算机科学 2025-09-30 Weiqiu You , Anton Xue , Shreya Havaldar , Delip Rao , Helen Jin , Chris Callison-Burch , Eric Wong

Early research into data poisoning attacks against Large Language Models (LLMs) demonstrated the ease with which backdoors could be injected. More recent LLMs add step-by-step reasoning, expanding the attack surface to include the…

密码学与安全 · 计算机科学 2025-09-09 Hanna Foerster , Ilia Shumailov , Yiren Zhao , Harsh Chaudhari , Jamie Hayes , Robert Mullins , Yarin Gal

Transferable backdoors pose a severe threat to the Pre-trained Language Models (PLMs) supply chain, yet defensive research remains nascent, primarily relying on detecting anomalies in the output feature space. We identify a critical flaw…

密码学与安全 · 计算机科学 2025-12-09 Tianhang Zhao , Wei Du , Haodong Zhao , Sufeng Duan , Gongshen Liu

Large language models (LLMs) have achieved remarkable success across diverse applications but remain vulnerable to jailbreak attacks, where attackers craft prompts that bypass safety alignment and elicit unsafe responses. Among existing…

计算与语言 · 计算机科学 2026-03-04 Zhi Xu , Jiaqi Li , Xiaotong Zhang , Hong Yu , Han Liu

Enabling large language models (LLMs) to solve complex reasoning tasks is a key step toward artificial general intelligence. Recent work augments LLMs with external tools to enable agentic reasoning, achieving high utility and efficiency in…

计算机视觉与模式识别 · 计算机科学 2026-01-27 Qi Li , Xinchao Wang

The rapid adoption of Large Language Model (LLM) agents and multi-agent systems enables remarkable capabilities in natural language processing and generation. However, these systems introduce security vulnerabilities that extend beyond…

密码学与安全 · 计算机科学 2026-05-12 Matteo Lupinacci , Francesco Aurelio Pironti , Francesco Blefari , Francesco Romeo , Luigi Arena , Angelo Furfaro

Backdoor unalignment attacks against Large Language Models (LLMs) enable the stealthy compromise of safety alignment using a hidden trigger while evading normal safety auditing. These attacks pose significant threats to the applications of…

密码学与安全 · 计算机科学 2025-06-23 Biao Yi , Tiansheng Huang , Sishuo Chen , Tong Li , Zheli Liu , Zhixuan Chu , Yiming Li

Diffusion Large Language Models (dLLMs) are rapidly emerging alongside autoregressive models as a powerful paradigm for complex reasoning, with reinforcement learning increasingly used for downstream alignment. Existing trajectory-based RL…

机器学习 · 计算机科学 2025-11-20 Ranfei Chen , Ming Chen , Kaifei Wang

Mitigating explicit and implicit biases in Large Language Models (LLMs) has become a critical focus in the field of natural language processing. However, many current methodologies evaluate scenarios in isolation, without considering the…

计算与语言 · 计算机科学 2025-02-04 Robert Morabito , Sangmitra Madhusudan , Tyler McDonald , Ali Emami

We introduce LADDER (Learning through Autonomous Difficulty-Driven Example Recursion), a framework which enables Large Language Models to autonomously improve their problem-solving capabilities through self-guided learning by recursively…

机器学习 · 计算机科学 2025-03-06 Toby Simonds , Akira Yoshiyama

The inherent capabilities of a language model (LM) and the reasoning strategies it employs jointly determine its performance in reasoning tasks. While test-time scaling is regarded as an effective approach to tackling complex reasoning…

计算与语言 · 计算机科学 2025-05-27 Zhihong Pan , Kai Zhang , Yuze Zhao , Yupeng Han

Current large-language models (LLMs) typically adopt a fixed reasoning strategy, either simple or complex, for all questions, regardless of their difficulty. This neglect of variation in task and reasoning process complexity leads to an…

计算与语言 · 计算机科学 2025-05-27 Yi Wang , Junxiao Liu , Shimao Zhang , Jiajun Chen , Shujian Huang

Large Vision-Language Models (LVLMs) have demonstrated remarkable capabilities across various tasks, yet they remain vulnerable to backdoor attacks. Existing defense methods predominantly focus on sample-level defense, which relies on the…

计算机视觉与模式识别 · 计算机科学 2026-05-18 Xuanyu Ge , Zhongqi Wang , Jie Zhang , Shiguang Shan , Xilin Chen