5 papers
AutoDojo: Adaptive Black-Box Attacks Reveal the Limits of IPI Defenses and Task-Specification Effects in LLM Agents
Xinhang Ma, Taoran Li, Chaowei Xiao +3
Indirect prompt injection (IPI) is a major security threat to LLM-powered agents. Thus, a growing body of work have proposed a variety of defensive approaches against IPI. These ca…
Layer-Targeted Multilingual Knowledge Erasure in Large Language Models
Taoran Li, Varun Chandrasekaran, Zhiyuan Yu
Recent work has demonstrated that machine unlearning in Large Language Models (LLMs) fails to generalize across languages: knowledge erased in one language frequently remains acces…
RAIN-Merging: A Gradient-Free Method to Enhance Instruction Following in Large Reasoning Models with Preserved Thinking Format
Zhehao Huang, Yuhang Liu, Baijiong Lin +5
Large reasoning models (LRMs) excel at a long chain of reasoning but often fail to faithfully follow instructions regarding output format, constraints, or specific requirements. We…
The Erasure Illusion: Stress-Testing the Generalization of LLM Forgetting Evaluation
Hengrui Jia, Taoran Li, Jonas Guan +1
Machine unlearning aims to remove specific data influences from trained models, a capability essential for adhering to copyright laws and ensuring AI safety. Current unlearning met…
SoK: Understanding (New) Security Issues Across AI4Code Use Cases
Qilong Wu, Taoran Li, Tianyang Zhou +1
AI-for-Code (AI4Code) systems are reshaping software engineering, with tools like GitHub Copilot accelerating code generation, translation, and vulnerability detection. Alongside t…