Defending Jailbreak Prompts via In-Context Adversarial Game

arXiv:2402.13148v1 Announce Type: new
Abstract: Large Language Models (LLMs) demonstrate remarkable capabilities across diverse applications. However, concerns regarding their security, particularly the vulnerability to jailbreak attacks, persist. Drawing inspiration from adversarial training in deep learning and LLM agent learning processes, we introduce the In-Context Adversarial Game (ICAG) for defending against jailbreaks without the need for fine-tuning. ICAG leverages agent learning to conduct an adversarial game, aiming to dynamically extend knowledge to defend against jailbreaks. Unlike traditional methods that rely on …

abstract adversarial adversarial training agent applications arxiv attacks capabilities concerns context cs.cr cs.lg deep learning diverse diverse applications game inspiration jailbreak language language models large language large language models llm llms processes prompts security training type via vulnerability