LCMP
TAMEing Long Contexts in Personalization: Towards Training-Free and State-Aware MLLM Personalized Assistant
Rongpei Hong, Jian Lang, Ting Zhong, Yong Wang, Fan Zhou · 2025-12-25
- Modality
- Image
- Data
- Synthetic
- Task · ground truth
- QA/Gen. · Ans.
- Scope
- Memory-specific benchmark
What it tests
- CECross-modal evidenceDirectly evaluated
- LELong-range evidenceDirectly evaluated
- SCState changesDirectly evaluated
- CFConflicting memoriesNot evaluated
- UQUnanswerable queriesNot evaluated
- MAMemory-grounded actionNot evaluated
- FGFine-grained visual detailPartially evaluated
- ETEntity trackingDirectly evaluated
- VCVisual state changeDirectly evaluated
Survey notes
Personalized visual concepts acquire changing short- and long-term attributes through historical dialogues. Later VQA probes test attribute retention and temporal validity.