Memory Intelligence Agent

2026-04-06

AI Summary

This work addresses the limitations of existing deep reasoning agents, which suffer from inefficient memory evolution and high storage and retrieval costs, hindering effective reasoning and autonomous self-improvement. To overcome these challenges, the authors propose the Memory Intelligence Agent (MIA) framework, featuring a Manager-Planner-Executor architecture that innovatively integrates parametric and non-parametric memory with a bidirectional conversion mechanism to enable online memory evolution during reasoning. The framework incorporates alternating reinforcement learning, test-time continual learning, memory compression, search-based planning, and a reflection module coupled with unsupervised judgment to significantly enhance autonomous evolution in open-ended environments. Evaluated across 11 benchmark tasks, the proposed method demonstrates substantial improvements over state-of-the-art approaches in both reasoning efficiency and self-evolution capabilities.

Abstract

Deep research agents (DRAs) integrate LLM reasoning with external tools. Memory systems enable DRAs to leverage historical experiences, which are essential for efficient reasoning and autonomous evolution. Existing methods rely on retrieving similar trajectories from memory to aid reasoning, while suffering from key limitations of ineffective memory evolution and increasing storage and retrieval costs. To address these problems, we propose a novel Memory Intelligence Agent (MIA) framework, consisting of a Manager-Planner-Executor architecture. Memory Manager is a non-parametric memory system that can store compressed historical search trajectories. Planner is a parametric memory agent that can produce search plans for questions. Executor is another agent that can search and analyze information guided by the search plan. To build the MIA framework, we first adopt an alternating reinforcement learning paradigm to enhance cooperation between the Planner and the Executor. Furthermore, we enable the Planner to continuously evolve during test-time learning, with updates performed on-the-fly alongside inference without interrupting the reasoning process. Additionally, we establish a bidirectional conversion loop between parametric and non-parametric memories to achieve efficient memory evolution. Finally, we incorporate a reflection and an unsupervised judgment mechanisms to boost reasoning and self-evolution in the open world. Extensive experiments across eleven benchmarks demonstrate the superiority of MIA.

Problem

• memory evolution
• storage cost
• retrieval cost
• deep research agents
• reasoning efficiency

Innovation

• Memory Intelligence Agent
• test-time learning
• parametric memory
• non-parametric memory
• alternating reinforcement learning