New white-box attack method targets LLMs using knowledge editing

Researchers have developed a new white-box attack method for large language models (LLMs) that leverages knowledge editing techniques. This approach modifies existing editing frameworks to incorporate associative knowledge retrieved directly from the model, allowing for attacks on broader thematic categories rather than just predefined prompts. Experiments show this method is more effective than previous techniques without significantly degrading the LLM's general performance. AI

IMPACT This research introduces a novel method for probing LLM vulnerabilities, potentially influencing future safety and security research.

RANK_REASON The cluster contains a research paper detailing a new method for attacking LLMs. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• IArxiv
• LLMs
• ScienceCast

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →