Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement Learning

TL;DR

Memory-R1 introduces a reinforcement learning framework that enables large language models to actively manage external memories, improving long-horizon reasoning and task performance with minimal supervision across multiple benchmarks.

Contribution

The paper presents a novel RL-based memory management system for LLMs, allowing dynamic and learned control over external memory operations, which was not addressed in prior static approaches.

Findings

Outperforms strong baselines with only 152 training QA pairs

Generalizes across diverse question types and benchmarks

Effective across multiple model scales (3B-14B)

Abstract

Large Language Models (LLMs) have demonstrated impressive capabilities across a wide range of NLP tasks, but they remain fundamentally stateless, constrained by limited context windows that hinder long-horizon reasoning. Recent efforts to address this limitation often augment LLMs with an external memory bank, yet most existing pipelines are static and heuristic-driven, lacking a learned mechanism for deciding what to store, update, or retrieve. We present Memory-R1, a reinforcement learning (RL) framework that equips LLMs with the ability to actively manage and utilize external memory through two specialized agents: a Memory Manager that learns structured operations, including ADD, UPDATE, DELETE, and NOOP; and an Answer Agent that pre-selects and reasons over relevant entries. Both agents are fine-tuned with outcome-driven RL (PPO and GRPO), enabling adaptive memory management with minimal supervision. With only 152 training QA pairs, Memory-R1 outperforms strong baselines and generalizes across diverse question types, three benchmarks (LoCoMo, MSC, LongMemEval), and multiple model scales (3B-14B).