Multimodal Agent MemorySubmit a paper

AniME

AniME: Adaptive Multi-Agent Planning for Long Animation Generation

Lisai Zhang, Baohan Xu, Siqian Yang, Mingyu Yin, Jing Liu, Chao Xu, et al. · 2025-10-10

Read the paperPDF

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
Video & streaming
Also: text, Image, audio
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. global production memory

    1. Represented asSpatial, temporal & cross-modal structured memoryCross-represented explicit
    2. Stored inHierarchical multi-tier store
    3. Retrieved byPlanner / tool-assisted retrieval
    4. Source evidenceSource linked & recoverable
    The director maintains global workflow and asset state used by downstream animation agents.Full paper: Method Details · PDF p. 3 · Figure 1

How memory changes (T2 operations)

  • Update · Rule-based update

    The director's fixed coordination workflow updates global production state as assets are created. For AniME, this specifically supports fixed-rule insertion or replacement of global production memory.

  • Revision · Rule-based state correction

    Production state is revised when downstream agents return updated assets and control conditions. For AniME, this specifically supports fixed-rule correction of stored state of global production memory.

Related systems