Multimodal Agent MemorySubmit a paper

OptimusVLA

Global Prior Meets Local Consistency: Dual-Memory Augmented Vision-Language-Action Model for Efficient Robotic Manipulation

Zaijing Li, Bing Hu, Rui Shao, Gongwei Chen, Dongmei Jiang, Pengwei Xie, et al. · 2026-05-18

Read the paperPDF

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
Embodied & 3D
Also: Image, 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. trajectory priors and recurrent action state

    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
    Global Prior Memory retrieves trajectories and Local Consistency Memory tracks executed actions.Full paper: 3.2. Global Prior Memory · PDF p. 4 · Figure 1

How memory changes (T2 operations)

  • Update · Learned update module

    A fixed Mamba module updates action-consistency state. For OptimusVLA, this specifically supports a trained but frozen update module of trajectory priors and recurrent action state.

Related systems