Multimodal Agent MemorySubmit a paper

ATMem

What Memory Do GUI Agents Really Need? From Passive Records to Active Task-Driving States

Chen Liu, Ling Chen, Hanzhang Zhou, Xu Zhang, Quyu Kong, Panrong Tong, et al. · 2026-07-02

Read the paperPDF

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
GUI & computer use
Also: text
Representation
Cross-represented explicit

How experience becomes memory

Each path is one memory component, traced from what is stored to how it is found again.

  1. active task-driving execution state

    1. Represented asSpatial, temporal & cross-modal structured memoryCross-represented explicit
    2. Stored inFlat sequential records
    3. Retrieved byMetadata / spatiotemporal retrieval
    4. Source evidenceSource linked & recoverable
    Task values are continually linked to their workflow role and current use status to drive action selection.Full-paper method / memory representation and retrieval · PDF p. 2 · Figure 1

How memory changes (T2 operations)

  • Update · Learned update module

    A learned but fixed deployed mechanism continually updates value roles and workflow status. For ATMem, this specifically supports a trained but frozen update module of active task-driving execution state.

  • Revision · Rule-based state correction

    The active state replaces outdated or already-used workflow values. For ATMem, this specifically supports fixed-rule correction of stored state of active task-driving execution state.

Related systems