4 papers
SAFETYEDIT: Sparse Internal Targets with Prompt-Wide Control for LLM Jailbreak Backdoors
Yutong Ke, Yansu Wang, Zeyu Zhang +3
Model-editing backdoors aim to induce harmful responses under a trigger while preserving ordinary behavior on untriggered input. Recent model-editing approaches target affirmative…
A Few Neurons Reveal When LLMs Misuse Tools: Sparse Detection and Selective Steering for Reliable Tool Use
Yutong Ke, Ming Yin, Chongwen Zhao +1
Agentic LLMs exhibit three consequential tool-use failures: invalid arguments (validity), unnecessary calls (over-calling), and omitted calls when tools are needed (missing). We fi…
SmoothGuard: Defending Multimodal Large Language Models with Noise Perturbation and Clustering Aggregation
Guangzhi Su, Shuchang Huang, Yutong Ke +3
Multimodal large language models (MLLMs) have achieved impressive performance across diverse tasks by jointly reasoning over textual and visual inputs. Despite their success, these…
Unraveling LLM Jailbreaks Through Safety Knowledge Neurons
Chongwen Zhao, Yutong Ke, Kaizhu Huang
Large Language Models (LLMs) are increasingly attracting attention in various applications. Nonetheless, there is a growing concern as some users attempt to exploit these models fo…