1 paper · 1 filter
Yu Li, Qikun Cai, Tao Huang +1
Memory-augmented and tool-using agents expose exact private values when remote LLMs process retrieved memory, tool actions, and intermediate observations. One-way masking limits di…