Multimodal Agent MemorySubmit a paper

Audio and Speech Memory

8 systems and 10 benchmarks whose primary modality is audio & speech.

ImageVideo & streamingGUI & computer useEmbodied & 3D

How they represent memory

How T2 systems change it

Representation × management tier

T1 Static Memory 1 · 13% T2 Dynamic Memory 7 · 88% T3 Self-Evolving Memory 0 · 0% Total
Source- and modality-specific – – – 0
Cross-represented explicit – – – 0
Internalized latent & parametric – 2 – 2
Hybrid 1 5 – 6

Each system counted once, at its highest tier. Circle area shows the value. Hover to compare with chance; click to list the systems.

What its benchmarks test

Directly evaluated, of 10 benchmarks in the survey's Table 2
  • Cross-modal evidenceCE6
  • Long-range evidenceLE8
  • State changesSC5
  • Conflicting memoriesCF3
  • Unanswerable queriesUQ2
  • Memory-grounded actionMA0
  • Speaker trackingSP4
  • Acoustic-cue retentionAC5
  • Repair persistenceRP1

Benchmarks

All 8 systems