DeltaMem: Towards Agentic Memory Management via Reinforcement Learning

TL;DR

DeltaMem is a reinforcement learning-based system that improves agentic memory management in conversational AI, outperforming existing methods across multiple long-term memory benchmarks.

Contribution

The paper introduces DeltaMem, a novel RL framework for persona-centric memory management, inspired by human memory evolution, with a new memory updating reward and dataset.

Findings

DeltaMem outperforms all baselines on LoCoMo, HaluMem, and PersonaMem benchmarks.

Both training-free and RL-trained DeltaMem show superior performance.

The approach effectively reduces information loss in complex multi-agent memory systems.

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.