Rethinking Memory as Continuously Evolving Connectivity

2026-05-27

AI Summary

This work addresses the limitations of conventional memory-augmented large language models, which treat memory as static storage and struggle to adapt to dynamic agent environments with evolving tasks and heterogeneous feedback. To overcome this, the authors propose FluxMem, a novel framework that models memory as a continuously evolving heterogeneous graph. Through three stages—initial graph construction, feedback-driven optimization, and long-term consolidation—FluxMem enables joint dynamic refinement of both memory content and topological structure. The approach incorporates a learnable graph evolution mechanism, leveraging heterogeneous graph neural networks, feedback-guided graph pruning and link repair, abstraction-level alignment, and distillation of successful trajectories. Evaluated on the LoCoMo, Mind2Web, and GAIA benchmarks, FluxMem achieves state-of-the-art performance, significantly enhancing agent adaptability and generalization in complex environments.

Abstract

Existing memory-augmented LLM agents often treat memory as a static repository with pre-defined representations and fixed retrieval pipelines, which is brittle in dynamic agentic environments where feedback, task variation, and heterogeneous signals continuously reshape what should be remembered and how it should be connected. To address this, we propose FluxMem, a connectivity-evolving memory framework that models memory as a heterogeneous graph and progressively refines its topology through three stages: initial connection formation, feedback-driven refinement, and long-term consolidation. During execution, FluxMem repairs missing links, prunes interference, aligns abstraction granularity, and distills recurrent successful trajectories into reusable procedural circuits, guided by one metric for memory generalizability and evolutionary maturity. Across three fundamentally distinct benchmarks including LoCoMo, Mind2Web, and GAIA, FluxMem achieves consistent state-of-the-art performance, demonstrating strong adaptation and generalization in complex agentic environments. The code will be open-sourced in https://github.com/zjunlp/LightMem.

Problem

• memory
• dynamic agentic environments
• heterogeneous signals
• task variation
• feedback

Innovation

• connectivity-evolving memory
• heterogeneous graph
• memory generalizability
• feedback-driven refinement
• procedural circuits