1
From Weak Cues to Real Identities: Evaluating Inference-Driven De-Anonymization in LLM Agents
大模型代理能从对话中的微小线索精确推断真实身份，隐私风险远超想象
arXiv:2603.18382v2 Announce Type: replace Abstract: Anonymization is often assumed to protect privacy once explicit identifiers are removed, because r…