Multimodal Agent MemorySubmit a paper

TRM-VLA

TRM-VLA: Temporal-Aware Chain-of-Thought Reasoning and Memorization for Vision-Language-Action Models

Xiang Li, Ya-Li Li, Yuan Wang, Shengjin Wang · 2026-06-01

Read the paperPDF

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
Embodied & 3D
Also: Image, text
Representation
Hybrid
Distinct operative representation paths are explicitly linked for downstream retrieval, reasoning, or control.

How experience becomes memory

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

  1. selected embodied reasoning trajectories

    1. Represented asPerception-to-text records & summariesCross-represented explicit
    2. Stored inFlat sequential records
    3. Retrieved byMetadata / spatiotemporal retrieval
    4. Source evidenceSource not retained
    Keyframe-triggered reasoning produces selected CoT records.Full paper: 3. Method · PDF p. 3 · Figure 1

  2. granularity-adaptive context state

    1. Represented asPersistent multimodal latent stateInternalized latent & parametric
    2. Stored inHierarchical multi-tier store
    3. Retrieved byLatent / parametric activation
    4. Source evidenceSource not retained
    Context memory stores and retrieves historical reasoning features.Full paper: 3. Method · PDF p. 3 · Figure 1

How memory changes (T2 operations)

  • Update · Learned update module

    Fixed GCM updates reasoning context across keyframes. For TRM-VLA, this specifically supports a trained but frozen update module of selected embodied reasoning trajectories.

  • Revision · Rule-based state correction

    During operation, TRM-VLA applies fixed-rule correction of stored state to selected embodied reasoning trajectories; the memory state changes while the governing update policy remains fixed.

Related systems