DeltaMem: Towards Agentic Memory Management via Reinforcement Learning

2026-04-01

AI Summary

This work addresses the challenges of information loss and cross-scenario fragility in existing role-based multi-agent dialogue systems, which struggle to effectively manage long-term memory. To this end, it reformulates role-centric memory modeling as a single-agent end-to-end task and introduces the first dialogue dataset annotated with operation-level memory update labels. The authors further propose a memory-aware Levenshtein distance as a reward signal for reinforcement learning to optimize memory management strategies. Experimental results demonstrate that the proposed approach significantly outperforms current production-level baselines across multiple long-term memory benchmarks—including LoCoMo, HaluMem, and PersonaMem—with substantial improvements in DeltaMem scores before and after training.

Abstract

Recent advances in persona-centric memory have revealed the powerful capability of multi-agent systems in managing persona memory, especially in conversational scenarios. However, these complex frameworks often suffer from information loss and are fragile across varying scenarios, resulting in suboptimal performance. In this paper, we propose DeltaMem, an agentic memory management system that formulates persona-centric memory management as an end-to-end task within a single-agent setting. To further improve the performance of our agentic memory manager, we draw inspiration from the evolution of human memory and synthesize a user-assistant dialogue dataset along with corresponding operation-level memory updating labels. Building on this, we introduce a novel Memory-based Levenshtein Distance to formalize the memory updating reward, and propose a tailored reinforcement learning framework to further enhance the management capabilities of DeltaMem. Extensive experiments show that both training-free and RL-trained DeltaMem outperform all product-level baselines across diverse long-term memory benchmarks, including LoCoMo, HaluMem, and PersonaMem.

Problem

• persona-centric memory
• memory management
• multi-agent systems
• information loss
• scenario robustness

Innovation

• agentic memory management
• reinforcement learning
• persona-centric memory
• Memory-based Levenshtein Distance
• memory updating