Semantic-TVM: Structure-Preserving Trustworthy Virtual Memory for Memory-Augmented and Tool-Using Agents

2026-09-14

AI Summary

为了解决远程模型处理内存和工具时暴露私有值的问题，提出了一种保持精确值本地化的可信虚拟内存方法，通过局部替换敏感片段来保护隐私并保持任务相关上下文。

Abstract

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 direct exposure but removes values needed for trusted execution and can leak them through later observations. We propose Trustworthy Virtual Memory (TVM), a closed-loop runtime that keeps exact-value state local while presenting a protected view to the remote model. Within this single runtime, Rule-TVM replaces whole protected fields with locally recoverable handles, and Semantic-TVM instead replaces only sensitive spans predicted by a trusted local model, preserving surrounding task-relevant context. On Memory-EHR and Memory-RAP across two providers, span-level projection recovers most of the EHR utility lost under whole-field replacement (Task Success 84.17% vs. 52.33% on DeepSeek) while measured exposure stays low and workflows remain executable.

Problem

• Memory-augmented agents
• Tool-using agents
• Private values exposure
• One-way masking
• Trusted execution

Innovation

• Trustworthy Virtual Memory
• Semantic-TVM
• memory-augmented agents
• sensitive spans
• closed-loop runtime