Multimodal Agent MemorySubmit a paper

Mem-W

Mem-W: Latent Memory-Native GUI Agents

Guibin Zhang, Yaohui Ling, Fanci Meng, Kun Wang, Shuicheng Yan · 2026-05-10

Read the paperPDF

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
GUI & computer use
Also: Image, text, actions
Representation
Internalized latent & parametric

How experience becomes memory

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

  1. compressed historical trajectories and expired in-session segments

    1. Represented asPersistent multimodal latent stateInternalized latent & parametric
    2. Stored inVector-indexed collection, Flat sequential records
    3. Retrieved bySimilarity search, Latent / parametric activation
    4. Source evidenceSource not retained
    A shared trajectory-to-latent compressor converts retrieved historical trajectories and expired current-session segments into compact tokens woven into the live policy context.Full paper: 3 Method · PDF p. 4 · Figure 1

How memory changes (T2 operations)

  • Update · Learned update module

    As the session advances, expired trajectory prefixes are compressed by the trained module into new latent working-memory blocks. For Mem-W, this specifically supports a trained but frozen update module of compressed historical trajectories and expired in-session segments.

Related systems