Learn to Disguise: Avoid Refusal Responses in LLM's Defense via a Multi-agent Attacker-Disguiser Game

arXiv:2404.02532v1 Announce Type: cross
Abstract: With the enhanced performance of large models on natural language processing tasks, potential moral and ethical issues of large models arise. There exist malicious attackers who induce large models to jailbreak and generate information containing illegal, privacy-invasive information through techniques such as prompt engineering. As a result, large models counter malicious attackers' attacks using techniques such as safety alignment. However, the strong defense mechanism of the large model through rejection replies is easily identified by …

abstract agent arxiv cs.ai cs.cl defense ethical game generate information jailbreak language language processing large models learn llm multi-agent natural natural language natural language processing performance privacy processing responses tasks through type via