Rethinking Memory Mechanisms of Foundation Agents in the Second Half: A Survey

TL;DR

This survey reviews recent advances in memory mechanisms for foundation agents, emphasizing their role in enabling long-term, dynamic, and user-dependent AI interactions in real-world environments.

Contribution

It provides a comprehensive, unified framework for understanding different memory types, mechanisms, and evaluation methods in foundation agents, highlighting open challenges and future research directions.

Findings

Memory is critical for long-horizon, dynamic AI tasks.

Various memory substrates and mechanisms are analyzed.

Evaluation benchmarks for memory utility are reviewed.

Abstract

The research of artificial intelligence is undergoing a paradigm shift from prioritizing model innovations over benchmark scores towards emphasizing problem definition and rigorous real-world evaluation. As the field enters the "second half," the central challenge becomes real utility in long-horizon, dynamic, and user-dependent environments, where agents face context explosion and must continuously accumulate, manage, and selectively reuse large volumes of information across extended interactions. Memory, with hundreds of papers released this year, therefore emerges as the critical solution to fill the utility gap. In this survey, we provide a unified view of foundation agent memory along three dimensions: memory substrate (internal and external), cognitive mechanism (episodic, semantic, sensory, working, and procedural), and memory subject (agent- and user-centric). We then analyze how memory is instantiated and operated under different agent topologies and highlight learning policies over memory operations. Finally, we review evaluation benchmarks and metrics for assessing memory utility, and outline various open challenges and future directions.