MemForm: Benchmarking Memory Formation Poisoning in LLM Agents
Factual evidence
What the source reports
Researchers introduced MemForm, a benchmark evaluating how memory formation interfaces in LLM agents are vulnerable to adversarial poisoning.
Inspect the evidence
- Inclusion basis
- Enterprise AI
- Publisher and source type
- arXiv cs.CL — Computation and Language · RESEARCH
- Published by source
- 6 November 2026
- Collected by OneBench
- 7 Nov 2026, 02:01 UK
- Original headline
- MemForm: Benchmarking Memory Formation Poisoning in LLM Agents ↗
Stored source excerpt
arXiv:2610.14431v2 Announce Type: replace-cross Abstract: Long-term memory enables LLM agents to learn from past interactions but exposes memory formation to adversarial influence. Existing studies primarily…
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 systems create security attack vectors where malicious inputs corrupt long-term decision-making capabilities.
Do what
Review agent memory architecture security controls with the team responsible for artificial intelligence safety and risk mitigation.