Saved in:
Bibliographic Details
Main Authors: Mi, Hongze, Feng, Yibo, Lu, WenJie, Cao, Song, Li, Jinyuan, Li, Yanming, Zhang, Xuelin, Luo, Haotian, Peng, Songyang, Cui, He, Tian, Tengfei, Fang, Jun, Chai, Hua, Tan, Naiqiang
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.22528
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908799120441344
author Mi, Hongze
Feng, Yibo
Lu, WenJie
Cao, Song
Li, Jinyuan
Li, Yanming
Zhang, Xuelin
Luo, Haotian
Peng, Songyang
Cui, He
Tian, Tengfei
Fang, Jun
Chai, Hua
Tan, Naiqiang
author_facet Mi, Hongze
Feng, Yibo
Lu, WenJie
Cao, Song
Li, Jinyuan
Li, Yanming
Zhang, Xuelin
Luo, Haotian
Peng, Songyang
Cui, He
Tian, Tengfei
Fang, Jun
Chai, Hua
Tan, Naiqiang
contents Multimodal Large Language Model (MLLM) agents facilitate Graphical User Interface (GUI) automation but struggle with long-horizon, cross-application tasks due to limited context windows. While memory systems provide a viable solution, existing paradigms struggle to adapt to dynamic GUI environments, suffering from a granularity mismatch between high-level intent and low-level execution, and context pollution where the static accumulation of outdated experiences drives agents into hallucination. To address these bottlenecks, we propose the Darwinian Memory System (DMS), a self-evolving architecture that constructs memory as a dynamic ecosystem governed by the law of survival of the fittest. DMS decomposes complex trajectories into independent, reusable units for compositional flexibility, and implements Utility-driven Natural Selection to track survival value, actively pruning suboptimal paths and inhibiting high-risk plans. This evolutionary pressure compels the agent to derive superior strategies. Extensive experiments on real-world multi-app benchmarks validate that DMS boosts general-purpose MLLMs without training costs or architectural overhead, achieving average gains of 18.0% in success rate and 33.9% in execution stability, while reducing task latency, establishing it as an effective self-evolving memory system for GUI tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2601_22528
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Darwinian Memory: A Training-Free Self-Regulating Memory System for GUI Agent Evolution
Mi, Hongze
Feng, Yibo
Lu, WenJie
Cao, Song
Li, Jinyuan
Li, Yanming
Zhang, Xuelin
Luo, Haotian
Peng, Songyang
Cui, He
Tian, Tengfei
Fang, Jun
Chai, Hua
Tan, Naiqiang
Artificial Intelligence
Multimodal Large Language Model (MLLM) agents facilitate Graphical User Interface (GUI) automation but struggle with long-horizon, cross-application tasks due to limited context windows. While memory systems provide a viable solution, existing paradigms struggle to adapt to dynamic GUI environments, suffering from a granularity mismatch between high-level intent and low-level execution, and context pollution where the static accumulation of outdated experiences drives agents into hallucination. To address these bottlenecks, we propose the Darwinian Memory System (DMS), a self-evolving architecture that constructs memory as a dynamic ecosystem governed by the law of survival of the fittest. DMS decomposes complex trajectories into independent, reusable units for compositional flexibility, and implements Utility-driven Natural Selection to track survival value, actively pruning suboptimal paths and inhibiting high-risk plans. This evolutionary pressure compels the agent to derive superior strategies. Extensive experiments on real-world multi-app benchmarks validate that DMS boosts general-purpose MLLMs without training costs or architectural overhead, achieving average gains of 18.0% in success rate and 33.9% in execution stability, while reducing task latency, establishing it as an effective self-evolving memory system for GUI tasks.
title Darwinian Memory: A Training-Free Self-Regulating Memory System for GUI Agent Evolution
topic Artificial Intelligence
url https://arxiv.org/abs/2601.22528