1
Natural Identifiers for Privacy and Data Audits in Large Language Models
用自然标识符给大模型做隐私审计，全新视角兼顾数据保护与可追溯性，AI安全领域值得关注的前沿研究。
arXiv:2606.24408v1 Announce Type: new Abstract: Assessing the privacy of large language models (LLMs) presents significant challenges. In particular, …