Multimodal Agent MemorySubmit a paper

HyLoVQA

HyLoVQA: Dynamic Hypernetwork-Generated Low-Rank Adaptation for Continual Visual Question Answering

Yiran Wang, Chenyi Xiong, Ziyue Qin, Miao Zhang, Kui Xiao, Zhifei Li · 2026-05-21

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
Image
Also: text
Representation
Hybrid
Distinct operative representation paths are deliberately linked or jointly consumed by downstream reasoning, generation, or control.

How experience becomes memory

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

  1. drift-resilient memory bank of visual-object and textual-task anchors

    1. Represented asPerception-to-text records & summariesCross-represented explicit
    2. Stored inFlat sequential records
    3. Retrieved bySimilarity search, Latent / parametric activation
    4. Source evidenceSource not retained
    Memory-bank anchor entries; hypernetwork-generated LoRA adapter weights. anchor retrieval conditions a hypernetwork that generates per-task LoRA adapters.Full paper: Update · PDF p. 3 · Figure 1 / method overview
  2. drift-resilient memory bank of visual-object and textual-task anchors

    1. Represented asExperience-conditioned parametric memoryInternalized latent & parametric
    2. Stored inModular adapter bank
    3. Retrieved bySimilarity search, Latent / parametric activation
    4. Source evidenceSource not retained
    Memory-bank anchor entries; hypernetwork-generated LoRA adapter weights. anchor retrieval conditions a hypernetwork that generates per-task LoRA adapters.Full paper: Update · PDF p. 3 · Figure 1 / method overview

How memory changes (T2 operations)

  • Update · Optimization-based update

    Retrieved visual-object and textual-task anchors are updated online through a fixed momentum-style procedure; the memory policy itself is not self-evolving. For HyLoVQA, this specifically supports a fixed operational optimization objective of drift-resilient memory bank of visual-object and textual-task anchors.

Related systems