MemoryDojo: Benchmarking Poisoning Through LLM Agent Memory
Factual evidence
What the source reports
Researchers introduced MemoryDojo, a benchmark evaluating how long-term memory poisoning in LLM agents affects multi-session safety.
Inspect the evidence
- Inclusion basis
- Enterprise AI
- Publisher and source type
- arXiv cs.CL — Computation and Language · RESEARCH
- Published by source
- 9 November 2026
- Collected by OneBench
- 10 Nov 2026, 02:01 UK
- Original headline
- MemoryDojo: Benchmarking Poisoning Through LLM Agent Memory ↗
Stored source excerpt
arXiv:2611.09721v1 Announce Type: cross Abstract: LLM agents benefit from using memory systems to improve performance on long-horizon, multi-session tasks. However, memory introduces a critical attack…
Short excerpt from the collected text, not the full source. Use the source link to read it in context.
The factual summary is a OneBench synthesis, not a quotation or independent verification. Collection time is not publication time. Open the source for its full context; related reporting can share the same underlying announcement.
OneBench interpretation
Institutional assessment
So what
Persistent agent memory creates cross-session attack vectors where unvalidated data retrieved from past sessions executes malicious actions.
Do what
Review input validation controls for agent memory persistence architectures with the team responsible for enterprise security.