ArtPrompt: ASCII Art-based Jailbreak Attacks against Aligned LLMs

arXiv:2402.11753v1 Announce Type: new
Abstract: Safety is critical to the usage of large language models (LLMs). Multiple techniques such as data filtering and supervised fine-tuning have been developed to strengthen LLM safety. However, currently known techniques presume that corpora used for safety alignment of LLMs are solely interpreted by semantics. This assumption, however, does not hold in real-world applications, which leads to severe vulnerabilities in LLMs. For example, users of forums often use ASCII art, a form of text-based art, …

abstract alignment art arxiv attacks cs.ai cs.cl data filtering fine-tuning interpreted jailbreak language language models large language large language models llm llms multiple safety semantics supervised fine-tuning type usage