2605.28773v1 May 27, 2026 cs.CL

Rethinking Memory as Continuously Evolving Connectivity

Zhixiang Wang
Zhixiang Wang
Citations: 2
h-index: 1
Huajun Chen
Huajun Chen
Citations: 4,759
h-index: 34
Xinle Deng
Xinle Deng
Citations: 141
h-index: 6
Guozhou Zheng
Guozhou Zheng
Citations: 535
h-index: 13
Baohua Dong
Baohua Dong
Citations: 66
h-index: 4
Hangcheng Zhu
Hangcheng Zhu
Citations: 44
h-index: 3
Ruohui Huang
Ruohui Huang
Citations: 60
h-index: 3
Gang Yu
Gang Yu
Citations: 13
h-index: 3
Buqiang Xu
Buqiang Xu
Citations: 15
h-index: 2
Jizhan Fang
Jizhan Fang
Citations: 124
h-index: 4
Haoliang Cao
Haoliang Cao
Citations: 0
h-index: 0
Ying Wei
Ying Wei
Citations: 6
h-index: 1
Feiyu Xiong
Feiyu Xiong
MemTensor (Shanghai) Technology Co., Ltd.
Citations: 674
h-index: 15
Haofen Wang
Haofen Wang
Citations: 280
h-index: 10
Ningyu Zhang
Ningyu Zhang
Citations: 127
h-index: 2

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.

0 Citations
0 Influential
70.876830454682 Altmetric
354.4 Score
Original PDF
875

No Analysis Report Yet

This paper hasn't been analyzed by Gemini yet.

Log in to request an AI analysis.

댓글

댓글을 작성하려면 로그인하세요.

아직 댓글이 없습니다. 첫 번째 댓글을 남겨보세요!