Attack and defense techniques in large language models: A survey and new perspectives

Zhiyu Liao, Kang Chen, Yuanguo Lin +5

Large Language Models (LLMs) have become central to numerous natural language processing tasks, but their vulnerabilities present significant security and ethical challenges. This…