6 papers
Calling the Bluff: Detecting Ever-Shifting Harmful Chat Dialogue via Ordered Reasoning Chain Regularization
Haojie Yu, Ziyou Jiang, Junjie Wang +4
Harmful chat dialogues are ever-shifting through type-shifting and lexical evasion, yet we find they share invariant principles, i.e., an Ordered Reasoning Chain (ORC) of recurring…
All Changes May Have Invariant Principles: Improving Ever-Shifting Harmful Meme Detection via Design Concept Reproduction
Ziyou Jiang, Mingyang Li, Junjie Wang +5
Harmful memes are ever-shifting in the Internet communities, which are difficult to analyze due to their type-shifting and temporal-evolving nature. Although these memes are shifti…
Know Thy Enemy: Securing LLMs Against Prompt Injection via Diverse Data Synthesis and Instruction-Level Chain-of-Thought Learning
Zhiyuan Chang, Mingyang Li, Yuekai Huang +6
Large language model (LLM)-integrated applications have become increasingly prevalent, yet face critical security vulnerabilities from prompt injection (PI) attacks. Defending agai…
Emerging from Ground: Addressing Intent Deviation in Tool-Using Agents via Deriving Real Calls into Virtual Trajectories
Qian Xiong, Yuekai Huang, Bo Yang +7
LLMs have advanced tool-using agents for real-world applications, yet they often lead to unexpected behaviors or results. Beyond obvious failures, the subtle issue of "intent devia…
Are All Prompt Components Value-Neutral? Understanding the Heterogeneous Adversarial Robustness of Dissected Prompt in Large Language Models
Yujia Zheng, Tianhao Li, Haotian Huang +8
Prompt-based adversarial attacks have become an effective means to assess the robustness of large language models (LLMs). However, existing approaches often treat prompts as monoli…
Butterfly Effects in Toolchains: A Comprehensive Analysis of Failed Parameter Filling in LLM Tool-Agent Systems
Qian Xiong, Yuekai Huang, Ziyou Jiang +4
The emergence of the tool agent paradigm has broadened the capability boundaries of the Large Language Model (LLM), enabling it to complete more complex tasks. However, the effecti…