Multimodal Agent MemorySubmit a paper

GCAgent

GCAgent: Graph-Centric Episodic Memory for Long Video Agents

Jeong Hun Yeo, Sangyun Chung, Sungjune Park, Dae Hoe Kim, Jinyoung Moon, Yong Man Ro · 2025-11-15

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
Video & streaming
Also: text
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. schematic episodic memory

    1. Represented asSpatial, temporal & cross-modal structured memoryCross-represented explicit
    2. Stored inGraph / relational store
    3. Retrieved byStructure traversal, Planner / tool-assisted retrieval
    4. Source evidenceSource linked & recoverable
    Schematic and narrative episodic memories encode entities, events, and causal-temporal relations.Full-paper method / memory representation and retrieval · PDF p. 2 · Figure 1
  2. narrative episode summaries

    1. Represented asPerception-to-text records & summariesCross-represented explicit
    2. Stored inHierarchical multi-tier store
    3. Retrieved bySimilarity search
    4. Source evidenceSource linked & recoverable
    Narrative memory summarizes episodes for later perception-action-reflection reasoning.Full-paper method / memory representation and retrieval · PDF p. 2 · Figure 1

How memory changes (T2 operations)

  • Consolidation · Episodic / hierarchical summarization

    Trajectory observations are consolidated into narrative episodes. For GCAgent, this specifically supports episodic or hierarchical summarization of schematic episodic memory.

  • Update · Prompted LLM/VLM record update

    During operation, GCAgent applies LLM-mediated memory editing to schematic episodic memory; the memory state changes while the governing update policy remains fixed.

Related systems