M3Eval
M$^3$Eval: Multi-Modal Memory Evaluation through Cognitively-Grounded Video Tasks
Jie Huang, Ruixun Liu, Sirui Sun, Xinyi Yang, Yin Li, Yixin Zhu, et al. · 2026-06-03
- Modality
- Video & streaming
- Data
- Hybrid (real + synthetic/simulated)
- Task · ground truth
- QA · Ans.+Time
- Scope
- Memory-specific benchmark
What it tests
- CECross-modal evidencePartially evaluated
- LELong-range evidenceDirectly evaluated
- SCState changesPartially evaluated
- CFConflicting memoriesNot evaluated
- UQUnanswerable queriesNot evaluated
- MAMemory-grounded actionNot evaluated
- TLTemporal localizationPartially evaluated
- TRTemporal relationsDirectly evaluated
- MEMulti-event integrationDirectly evaluated
Survey notes
Controlled sequence construction tests interference, delayed recall and false-memory behavior.