Search papers, labs, and topics across Lattice.
This paper introduces MemTxn, a governance layer designed to enhance the reliability of updates and recovery in persistent memory systems for large language model agents. By implementing a transaction boundary that verifies updates against their sources and manages conflicting information, MemTxn ensures that errors in writable memory do not compromise future behavior. The system demonstrates superior performance, achieving the highest average F1 score across various configurations and significantly outperforming existing methods in fault recovery and state restoration tasks.
MemTxn not only prevents memory corruption in long-running agents but also restores complete state after faults, achieving a remarkable 24-point improvement over existing systems.
Persistent memory lets long-running large language model agents reuse information across sessions and tasks. Yet errors in writable memory can persist and corrupt future behavior. Existing systems improve storage and retrieval, but they do not provide a transaction boundary for reliable updates and recovery. We therefore propose MemTxn, a governance layer outside the answer model. MemTxn verifies whether an update is supported by its source. It also selects the visible version when facts conflict and restores the application-visible state after a fault. The system uses Ordered PatchTest to validate writes, a Temporal Resolver to select versions, and a durable snapshot journal to recover state. On an item-disjoint audit, MemTxn accepts all 60 supported originals and rejects all 179 hard negatives. Under persistent multi-key faults on LongMemEval-S and LoCoMo states, it restores the complete declared active map without knowing the actual physical write set. On MemoryAgentBench FactConsolidation, MemTxn achieves the highest average F1 across all twelve answer-model configurations. It outperforms Dense by 17.06--24.07 points in five representative settings.
