Multimodal Agent MemorySubmit a paper

Darwinian Memory

Darwinian Memory: A Training-Free Self-Regulating Memory System for GUI Agent Evolution

Hongze Mi, Yibo Feng, WenJie Lu, Song Cao, Jinyuan Li, Yanming Li, et al. · 2026-01-30

Read the paperPDF

Management
T2 Dynamic Memory
Memory changes during deployment through a stable management policy.
Primary modality
GUI & computer use
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. reusable trajectory units

    1. Represented asPerception-to-text records & summariesCross-represented explicit
    2. Stored inHierarchical multi-tier store
    3. Retrieved bySimilarity search
    4. Source evidenceSource linked & recoverable
    Complex trajectories are decomposed into reusable units for compositional GUI execution.Full paper: 3. Method · PDF p. 2 · Figure 1

  2. survival-value ecosystem

    1. Represented asSpatial, temporal & cross-modal structured memoryCross-represented explicit
    2. Stored inGraph / relational store
    3. Retrieved byStructure traversal
    4. Source evidenceSource linked & recoverable
    Each unit carries utility and risk state that controls future survival, pruning, and plan inhibition.Full paper: Memory · PDF p. 4 · Figure 1

How memory changes (T2 operations)

  • Revision · Evidence-weighted belief revision

    Observed utility updates the survival values of reusable memory units. For Darwinian Memory, this specifically supports evidence-weighted revision of stored belief or confidence of reusable trajectory units.

  • Forgetting · Discrete eviction

    Low-value trajectories are actively pruned and high-risk plans inhibited. For Darwinian Memory, this specifically supports explicit deletion or eviction of stored items of reusable trajectory units.

  • Forgetting · Progressive decay

    During operation, Darwinian Memory applies progressive decay of stored importance or retention to reusable trajectory units; the memory state changes while the governing update policy remains fixed.

  • Revision · Rule-based state correction

    During operation, Darwinian Memory applies fixed-rule correction of stored state to reusable trajectory units; the memory state changes while the governing update policy remains fixed.

Related systems