跳到正文
arXiv cs.CL· David H. Yang, Blake Bullwinkel, Amanda Minnich, Mohammad Mohammadi Amiri, Giorgio Severi·· 11 小时前AI 评分45

MemoryDojo:基准测试 LLM 智能体记忆投毒

MemoryDojo: Benchmarking Poisoning Through LLM Agent Memory

AI 导读

研究者推出 MemoryDojo 基准,用于追踪投毒信息在 LLM 智能体记忆中的生命周期并测量其对后续行为的影响。该基准包含覆盖四个领域的 155 条多会话长程轨迹,支持五种记忆系统,评测发现所有记忆系统均可被记忆投毒,且下游影响差异显著。团队还提出记忆感知的投毒优化流程,可提升投毒效果并凸显不同记忆架构的差异。

正文 · 原文

View PDF HTML (experimental)

Abstract:LLM agents benefit from using memory systems to improve performance on long-horizon, multi-session tasks. However, memory introduces a critical attack surface. Malicious instructions encountered in one session may be stored and propagated through memory before being retrieved and acted upon in later sessions. Because memory systems differ in how they process and retrieve information, the downstream effects of poisoning can vary significantly across systems. To quantify the effects of poisoning across different memory architectures, we introduce MemoryDojo, a benchmark for tracing the lifecycle of poisoned information through agent memory and measuring its influence on subsequent agent behavior. MemoryDojo is easily extensible, consists of 155 multi-session long-horizon trajectories covering four domains, and supports five different memory systems. Through our evaluation, we find that all memory systems are susceptible to memory poisoning, with substantial variation in downstream effects. We further propose a memory-aware poison optimization process that increases the poisoning efficacy, highlighting differences between memory architectures. We hope that MemoryDojo will support research on understanding and mitigating agentic memory risks.
Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
Cite as: arXiv:2611.09721 [cs.CR]
  (or arXiv:2611.09721v1 [cs.CR] for this version)
  https://doi.org/10.48550/arXiv.2611.09721

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: David Hong Yang [view email]
[v1] Fri, 6 Nov 2026 17:00:09 UTC (418 KB)

来源:arXiv cs.CL · arxiv.org