Multimodal Agent MemorySubmit a paper

CANVAS

CANVAS: Continuity-Aware Narratives via Visual Agentic Storyboarding

Ishani Mondal, Yiwen Song, Mihir Parmar, Palash Goyal, Jordan Boyd-Graber, Tomas Pfister, et al. · 2026-04-15

Read the paperPDF

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
Video & streaming
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. World-state anchor memory for cross-shot continuity

    1. Represented asSpatial, temporal & cross-modal structured memoryCross-represented explicit
    2. Stored inFlat sequential records
    3. Retrieved byPlanner / tool-assisted retrieval
    4. Source evidenceSource not retained
    Storyboard state; Reference-set anchor frames. Reference-set anchors condition I2V; world state is consulted per shot.Full-paper method / memory representation and retrieval · PDF p. 20 · Figure 1 / method overview

How memory changes (T2 operations)

  • Update · Rule-based update

    World state is updated across shots by fixed agent rules to maintain continuity. For CANVAS, this specifically supports fixed-rule insertion or replacement of World-state anchor memory for cross-shot continuity.

Related systems