Multimodal Agent MemorySubmit a paper

L2-VMAS

Dual Latent Memory for Visual Multi-agent System

Xinlei Yu, Chengming Xu, Zhangquan Chen, Bo Yin, Cheng Yang, Yongbo He, et al. · 2026-01-31

Read the paperPDF

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
Image
Also: text
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. perception and thinking latent units

    1. Represented asPersistent multimodal latent stateInternalized latent & parametric
    2. Stored inVector-indexed collection
    3. Retrieved byLatent / parametric activation
    4. Source evidenceSource not retained
    Dual latent perception and thinking memories are synthesized and accessed on demand.Full paper: 3.2. Memory Synthesis · PDF p. 4 · Figure 1

How memory changes (T2 operations)

  • Update · Learned update module

    Fixed synthesis modules update latent memory across turns. For L2-VMAS, this specifically supports a trained but frozen update module of perception and thinking latent units.

  • Consolidation · Merging & deduplication

    Retrieved units are filtered and merged before injection. For L2-VMAS, this specifically supports merging or deduplication of overlapping entries of perception and thinking latent units.

Related systems