Multimodal Agent MemorySubmit a paper

POLAR

Personalizing Embodied Multimodal Large Language Model Agents over Long-term User Interactions

Jeongeun Lee, Chanyoung Park, Dongha Lee · 2026-05-25

Read the paperPDF

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
Embodied & 3D
Also: text, Image
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. personalized semantic context

    1. Represented asSpatial, temporal & cross-modal structured memoryCross-represented explicit
    2. Stored inGraph / relational store
    3. Retrieved bySimilarity search, Structure traversal
    4. Source evidenceSource not retained
    Multimodal knowledge-graph entries for semantic memory and episodic embodied memory. retrieves relevant memories to interpret implicit personalized requests and guide task execution.Full paper: Retrieved Memory · PDF p. 4 · Figure 1 / method overview

How memory changes (T2 operations)

  • Update · Prompted LLM/VLM record update

    During operation, POLAR applies LLM-mediated memory editing to personalized semantic context; the memory state changes while the governing update policy remains fixed.

  • Consolidation · Episodic / hierarchical summarization

    During operation, POLAR applies episodic or hierarchical summarization to personalized semantic context; the memory state changes while the governing update policy remains fixed.

  • Consolidation · Experience abstraction

    During operation, POLAR applies abstraction of episodes into reusable experience to personalized semantic context; the memory state changes while the governing update policy remains fixed.

Related systems