RESEARCH · RESEARCH · #2470
PerMIT: machine-checked integrity theorems and a trust-gated memory architecture for LLM agents
PerMIT is a memory architecture for LLM agents that enforces a strict trust model—only user-typed inputs and values confirmed at a gate are trusted—and mediates untrusted environment-derived values via a finite interface. The authors prove two integrity theorems (machine-checked in Lean 4), report zero observed attack success on Trojan Hippo with self- and lightweight verifiers, and simulate communities (up to 90 agents) on τ^2-bench showing scalable community-confirmation of memory and improved downstream agent performance.
KEY POINTS
- PerMIT is a memory architecture for LLM agents that enforces a strict trust model—only user-typed inputs and values confirmed at a gate are trusted—and mediates untrusted environment-derived values via a finite interface.
- The authors prove two integrity theorems (machine-checked in Lean 4), report zero observed attack success on Trojan Hippo with self- and lightweight verifiers, and simulate communities (up to 90 agents) on τ^2-bench showing scalable community-confirmation of memory and improved downstream agent performance.
- Provides a provable, machine-checked approach to securing persistent memory for LLM agents and a practical path for scalable, community-backed memory confirmation, which could reduce poisoning risks and enable safer shared or self-improving agents.
WHY IT MATTERS
Provides a provable, machine-checked approach to securing persistent memory for LLM agents and a practical path for scalable, community-backed memory confirmation, which could reduce poisoning risks and enable safer shared or self-improving agents.