Multimodal Agent MemorySubmit a paper

GA-VLN

GA-VLN: Geometry-Aware BEV Representation for Efficient Vision-Language Navigation

Jiahao Yang, Zihan Wang, Xiangyang Li, Xing Zhu, Yujun Shen, Yinghao Xu, et al. · 2026-05-21

Read the paperPDF

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
Embodied & 3D
Also: text, Image
Representation
Source- and modality-specific

How experience becomes memory

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

  1. agent-centric Bird's-Eye-View (BEV) spatial representations carrying geometry-aware features

    1. Represented asModality-specific encodings & geometric stateSource- and modality-specific
    2. Stored inFlat sequential records
    3. Retrieved byStructure traversal
    4. Source evidenceRetained but not accessible to the agent
    BEV map entries representing the local spatial structure with geometry-aware encoding. navigation policy reads geometry-aware BEV maps for action selection.Full-paper method / memory representation and retrieval · PDF p. 2 · Figure 1 / method overview

How memory changes (T2 operations)

  • Update · Rule-based update

    BEV map updates as the agent moves through the environment. For GA-VLN, this specifically supports fixed-rule insertion or replacement of agent-centric Bird's-Eye-View (BEV) spatial representations carrying geometry-aware features.

Related systems