A formal, graph-native memory architecture that treats agent memory as a versioned asset, dramatically outperforming Gemini 2.5 Pro on complex recall.
March 19, 2026
Original Paper
Graph-Native Cognitive Memory for AI Agents: Formal Belief Revision Semantics for Versioned Memory Architectures
arXiv · 2603.17244
The Takeaway
It grounds agent memory in formal belief revision semantics (AGM postulates) rather than just vector search. The architecture handles immutable revisions and mutable tags, solving long-standing issues with agent consistency and adversarial refusal.
From the abstract
While individual components for AI agent memory exist in prior systems, their architectural synthesis and formal grounding remain underexplored. We present Kumiho, a graph-native cognitive memory architecture grounded in formal belief revision semantics. The structural primitives required for cognitive memory -- immutable revisions, mutable tag pointers, typed dependency edges, URI-based addressing -- are identical to those required for managing agent-produced work as versionable assets, enablin