Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AriadneMem: Threading the Maze of Lifelong Memory for LLM Agents

Wenhui Zhu, Xiwen Chen, Zhipeng Wang, Jingjing Wang, Xuanzhao Dong, Minzhou Huang, Rui Cai, Hejian Sang, Hao Wang, Peijie Qiu, Yueyue Deng, Prayag Tiwari, Brendan Hogan Rappazzo, Yalin Wang

TL;DR

AriadneMem is proposed, a structured memory system that addresses failure modes via a decoupled two-phase pipeline and reduces the total runtime by 77.8% by offloading reasoning to the graph layer.

Abstract

Long-horizon LLM agents require memory systems that remain accurate under fixed context budgets. However, existing systems struggle with two persistent challenges in long-term dialogue: (i) \textbf{disconnected evidence}, where multi-hop answers require linking facts distributed across time, and (ii) \textbf{state updates}, where evolving information (e.g., schedule changes) creates conflicts with older static logs. We propose AriadneMem, a structured memory system that addresses these failure modes via a decoupled two-phase pipeline. In the \textbf{offline construction phase}, AriadneMem employs \emph{entropy-aware gating} to filter noise and low-information message before LLM extraction and applies \emph{conflict-aware coarsening} to merge static duplicates while preserving state transitions as temporal edges. In the \textbf{online reasoning phase}, rather than relying on expensive iterative planning, AriadneMem executes \emph{algorithmic bridge discovery} to reconstruct missing logical paths between retrieved facts, followed by \emph{single-call topology-aware synthesis}. On LoCoMo experiments with GPT-4o, AriadneMem improves \textbf{Multi-Hop F1 by 15.2\%} and \textbf{Average F1 by 9.0\%} over strong baselines. Crucially, by offloading reasoning to the graph layer, AriadneMem reduces \textbf{total runtime by 77.8\%} using only \textbf{497} context tokens. The code is available at https://github.com/LLM-VLM-GSL/AriadneMem.

AriadneMem: Threading the Maze of Lifelong Memory for LLM Agents

TL;DR

AriadneMem is proposed, a structured memory system that addresses failure modes via a decoupled two-phase pipeline and reduces the total runtime by 77.8% by offloading reasoning to the graph layer.

Abstract

Long-horizon LLM agents require memory systems that remain accurate under fixed context budgets. However, existing systems struggle with two persistent challenges in long-term dialogue: (i) \textbf{disconnected evidence}, where multi-hop answers require linking facts distributed across time, and (ii) \textbf{state updates}, where evolving information (e.g., schedule changes) creates conflicts with older static logs. We propose AriadneMem, a structured memory system that addresses these failure modes via a decoupled two-phase pipeline. In the \textbf{offline construction phase}, AriadneMem employs \emph{entropy-aware gating} to filter noise and low-information message before LLM extraction and applies \emph{conflict-aware coarsening} to merge static duplicates while preserving state transitions as temporal edges. In the \textbf{online reasoning phase}, rather than relying on expensive iterative planning, AriadneMem executes \emph{algorithmic bridge discovery} to reconstruct missing logical paths between retrieved facts, followed by \emph{single-call topology-aware synthesis}. On LoCoMo experiments with GPT-4o, AriadneMem improves \textbf{Multi-Hop F1 by 15.2\%} and \textbf{Average F1 by 9.0\%} over strong baselines. Crucially, by offloading reasoning to the graph layer, AriadneMem reduces \textbf{total runtime by 77.8\%} using only \textbf{497} context tokens. The code is available at https://github.com/LLM-VLM-GSL/AriadneMem.
Paper Structure (22 sections, 12 equations, 3 figures, 4 tables, 1 algorithm)

This paper contains 22 sections, 12 equations, 3 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Method
  3. Problem Setup & Pipeline Overview
  4. Phase I: Asynchronous Memory Construction
  5. Entropy-Aware Gating.
  6. Atomic Entry Extraction.
  7. Conflict-Aware Graph Coarsening.
  8. Outcome: Graph as a Navigational Prior.
  9. Phase II: Real-Time Structural Reasoning
  10. Fast Paths (Heuristic Short-Circuiting).
  11. Hybrid Retrieval.
  12. Base Graph Construction.
  13. Algorithmic Bridge Discovery (Steiner Tree Approximation).
  14. Multi-Hop Path Mining & Node Budget.
  15. Topology-Aware Reasoning.
  16. ...and 7 more sections

Figures (3)

  • Figure 1: Efficiency-Performance Trade-off on LoCoMo benchmark.
  • Figure 2: Overview of the AriadneMem architecture. The pipeline is decoupled into two phases: (I) Offline Memory Construction, which maintains an evolutionary graph via entropy-aware gating and conflict-aware coarsening to resolve state updates; and (II) Online Structural Reasoning, which connects disjoint evidence through algorithmic bridge discovery and performs topology-aware synthesis.
  • Figure 3: Qualitative Example of Structural Reasoning. A sample output showing how AriadneMem retrieves and serializes a coherent, timestamped evidence chain to answer a multi-hop question.