BMAM: Brain-inspired Multi-Agent Memory Framework

TL;DR

BMAM introduces a brain-inspired, multi-component memory system for language agents, significantly improving long-horizon reasoning and memory retention by mimicking cognitive memory structures.

Contribution

It proposes a novel, multi-component memory architecture inspired by cognitive systems, enhancing temporal reasoning and memory management in language-model-based agents.

Findings

Achieves 78.45% accuracy on LoCoMo benchmark

Episodic memory subsystem is critical for temporal reasoning

Organizes memories along explicit timelines for better retrieval

Abstract

Language-model-based agents operating over extended interaction horizons face persistent challenges in preserving temporally grounded information and maintaining behavioral consistency across sessions, a failure mode we term soul erosion. We present BMAM (Brain-inspired Multi-Agent Memory), a general-purpose memory architecture that models agent memory as a set of functionally specialized subsystems rather than a single unstructured store. Inspired by cognitive memory systems, BMAM decomposes memory into episodic, semantic, salience-aware, and control-oriented components that operate at complementary time scales. To support long-horizon reasoning, BMAM organizes episodic memories along explicit timelines and retrieves evidence by fusing multiple complementary signals. Experiments on the LoCoMo benchmark show that BMAM achieves 78.45 percent accuracy under the standard long-horizon evaluation setting, and ablation analyses confirm that the hippocampus-inspired episodic memory subsystem plays a critical role in temporal reasoning.