MULTIVERSE: Exposing Large Language Model Alignment Problems in Diverse Worlds

Large Language Model (LLM) alignment aims to ensure that LLM outputs match with human values. Researchers have demonstrated the severity of alignment problems with a large spectrum of jailbreak techniques that can induce LLMs to produce malicious content during conversations. Finding the corresponding jailbreaking prompts usually requires substantial human intelligence or computation resources. In this paper, we report that LLMs have different levels of alignment in various contexts. As such, by systematically constructing many contexts, called worlds, leveraging a Domain …

alignment conversations cs.cl cs.lg diverse human human intelligence intelligence jailbreak jailbreaking language language model large language large language model llm llms malicious content match multiverse prompts researchers spectrum values