CarMem (text-only)
CarMem: Enhancing Long-Term Memory in LLM Voice Assistants through Category-Bounding
Johannes Kirmayr, Lukas Stappen, Phillip Schneider, Florian Matthes, Elisabeth Andre · 2025-01
- Modality
- Audio & speech
- Data
- Synthetic
- Task · ground truth
- Ret./Learn · Pref.+Ret.
- Scope
- text only comparator
What it tests
- CECross-modal evidenceNot evaluated
- LELong-range evidenceDirectly evaluated
- SCState changesDirectly evaluated
- CFConflicting memoriesDirectly evaluated
- UQUnanswerable queriesNot evaluated
- MAMemory-grounded actionNot evaluated
- SPSpeaker trackingNot evaluated
- ACAcoustic-cue retentionNot evaluated
- RPRepair persistencePartially evaluated
Survey notes
A synthetic text-conversation protocol directly measures preference extraction, retrieval and maintenance.