John T. Halloran

This study compares two methods of safety unalignment (Jailbreak-Tuning and Weight Orthogonalization) across six LLMs and finds that Weight Orthogonalization (WO) significantly enhances malicious capabilities, making it a greater risk than Jailbreak-Tuning (JT).

The paper introduces RAG-Pref, a novel, training-free Retrieval Augmented Generation (RAG) method for preference alignment that significantly improves LLM refusal guardrails against agentic attacks with minimal computational overhead.

The paper proposes Open-Book Benign Rewriting (OBBR), a novel defense mechanism that uses LLM rewriting with benign samples to neutralize data poisoning attacks against LLMs, significantly improving safety performance.