TabPATE: Differentially Private Tabular In-Context Learning Without Public Data
TabPATE: 无需公开数据的差分隐私表格上下文学习
机构 * CISPA(CISPA亥姆霍兹信息安全中心) ; Anthropic ; Layer 6 AI
AI总结 提出TabPATE,一种无需公开数据的差分隐私PATE风格防御方法,通过教师模型分区和私有聚合生成学生上下文,在表格基准上保持可用性并将成员推断降至随机水平。
Comments Presented at the 2nd ICML Workshop on Foundation Models for Structured Data (2026)