G. Edward Suh

Enforcing Instruction Hierarchy via Augmented Intermediate Representations

How much can language models memorize?

Privasis: Synthesizing the Largest "Public" Private Dataset from Scratch

ReasAlign: Reasoning Enhanced Safety Alignment against Prompt Injection Attack

Mitigating Indirect Prompt Injection via Instruction-Following Intent Analysis

Efficient Metric for Distinguishing Memorization from Generalization in Large Language Models

AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMs

Trustworthy Machine Learning through Data-Specific Indistinguishability