SYNAPSE: Empowering LLM Agents with Episodic-Semantic Memory via Spreading Activation

TL;DR

Synapse introduces a dynamic, graph-based memory system for LLMs that enhances long-term reasoning by mimicking cognitive processes like spreading activation and temporal decay, leading to improved performance on complex tasks.

Contribution

This paper presents Synapse, a novel memory architecture for LLMs that models episodic-semantic memory as a dynamic graph with spreading activation, surpassing static retrieval methods.

Findings

Outperforms state-of-the-art in complex reasoning tasks

Effectively filters interference through lateral inhibition and decay

Enhances long-term memory retrieval in LLMs

Abstract

While Large Language Models (LLMs) excel at generalized reasoning, standard retrieval-augmented approaches fail to address the disconnected nature of long-term agentic memory. To bridge this gap, we introduce Synapse (Synergistic Associative Processing Semantic Encoding), a unified memory architecture that transcends static vector similarity. Drawing from cognitive science, Synapse models memory as a dynamic graph where relevance emerges from spreading activation rather than pre-computed links. By integrating lateral inhibition and temporal decay, the system dynamically highlights relevant sub-graphs while filtering interference. We implement a Triple Hybrid Retrieval strategy that fuses geometric embeddings with activation-based graph traversal. Comprehensive evaluations on the LoCoMo benchmark show that Synapse significantly outperforms state-of-the-art methods in complex temporal and multi-hop reasoning tasks, offering a robust solution to the "Contextual Tunneling" problem. Our code and data will be made publicly available upon acceptance.