MEMOREPAIR: Barrier-First Cascade Repair in Agentic Memory

2026-05-08

AI Summary

This work addresses the problem of cascading error propagation in agent memory caused by failures of source components. It formally characterizes this cascading update challenge and introduces a barrier-first cascading repair mechanism. The approach first retracts affected derived states and then constructs and validates new states based on valid predecessors and current interfaces, ensuring that only closure-consistent successors are published. The key innovation lies in a barrier-first repair contract and the formulation of repair selection as a maximum-weight predecessor closure problem, which admits an exact solution. By integrating impact tracing with an s-t min-cut algorithm, the method achieves efficient repair. Experiments on ToolBench and MemoryArena demonstrate that invalid memory exposure is reduced to 0%, while retaining 91.1–94.3% of valid successors and reducing repair cost to 0.57–0.76× that of full re-computation.

Abstract

Agentic memory evolves across tasks into durable derived artifacts: summaries, cached outputs, embeddings, learned skills, and executable tool procedures. When a source artifact is deleted, corrected, or invalidated by tool or API migration, descendants derived from that source can remain visible and steer future actions with stale support. We formalize this failure mode as the cascade update problem, where repair targets the visible derived state of the memory store. We present MemoRepair, a barrier-first cascade-repair contract for agentic memory. A repair event induces a controlled transition from invalidated descendant state to validated successor state: affected descendants are withdrawn before repair, successors are constructed from retained support and staged repaired predecessors under the current interface, and republication is restricted to validated predecessor-closed successors. This contract induces a scalarized repair-selection problem for a fixed repair-cost tradeoff. We show that the induced publication problem reduces to maximum-weight predecessor closure and can be solved exactly by a single s-t min-cut. Experiments on ToolBench and MemoryArena show that, with complete influence provenance, MemoRepair reduces invalidated-memory exposure from 69.8-94.3% under systems without cascade repair to 0%. Compared with exhaustive Repair all, it recovers 91.1-94.3% of validated successors while reducing normalized repair-operator cost from 1.00 to 0.57-0.76.

Problem

• agentic memory
• cascade update problem
• memory repair
• derived artifacts
• invalidated state

Innovation

• cascade repair
• agentic memory
• predecessor closure
• min-cut optimization
• memory provenance