Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Hanqi Jiang, Junhao Chen, Yi Pan, Ling Chen, Weihang You, Yifan Zhou, Ruidong Zhang, Yohannes Abate, Tianming Liu

TL;DR

This work introduces Synapse (Synergistic Associative Processing Semantic Encoding), a unified memory architecture that transcends static vector similarity and models memory as a dynamic graph where relevance emerges from spreading activation rather than pre-computed links.

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.

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

TL;DR

This work introduces Synapse (Synergistic Associative Processing Semantic Encoding), a unified memory architecture that transcends static vector similarity and models memory as a dynamic graph where relevance emerges from spreading activation rather than pre-computed links.

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.
Paper Structure (69 sections, 6 equations, 4 figures, 12 tables, 1 algorithm)

This paper contains 69 sections, 6 equations, 4 figures, 12 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Memory Allocation Capabilities
  4. Graph-Based and Structured Memory
  5. Semantic Similarity and Relational Retrieval
  6. Methodology
  7. Unified Episodic-Semantic Graph
  8. Node Construction.
  9. Topology.
  10. Graph Maintenance and Scalability.
  11. Cognitive Dynamics: Spreading Activation
  12. Initialization.
  13. Propagation with Fan Effect.
  14. Lateral Inhibition.
  15. Sigmoid Activation.
  16. ...and 54 more sections

Figures (4)

  • Figure 1: Overview of the Synapse architecture. (Left) A user query regarding "that guy from the ski trip" activates the graph via Dual Triggers: Lexical matching targets explicit entities ("Kendall"), while Semantic embedding targets implicit concepts ("Ski Trip"). (Center)Spreading Activation dynamically propagates relevance through the Unified Episodic-Semantic Graph. Note how the bridge node "Mark" (purple) is activated despite not appearing in the query, connecting the disjoint concepts of "Ski Trip" and "Dating". (Right) The Triple Hybrid Scoring layer reranks candidates, successfully retrieving the ground truth ("broke up with Mark") while suppressing semantically similar but logically irrelevant distractors ("going skiing") via lateral inhibition.
  • Figure 2: Sensitivity analysis of Top-$k$ retrieval on LoCoMo benchmark. Performance is robust across $k \in [20, 40]$, with optimal stability around $k=30$. Star markers denote A-Mem baseline performance at their experiment settings.
  • Figure 3: Prompt template for extracting semantic nodes and edges. The prompt enforces a strict "Reason-then-Extract" workflow (CoT) and categorizes memories into specific cognitive types to structure the graph effectively.
  • Figure 4: Cognitive Tunneling: Lateral inhibition aggressively prunes low-degree details in the presence of highly activated hubs, leading to loss of "minor" facts.