SpaMEM
SpaMEM: Benchmarking Dynamic Spatial Reasoning via Perception-Memory Integration in Embodied Environments
Chih-Ting Liao, Xi Xiao, Chunlei Meng, Zhangquan Chen, Yitong Qiao, Weilin Zhou, et al. · 2026-05-29
- Modality
- Embodied & 3D
- Data
- Simulated
- Task · ground truth
- QA/Act. · Ans.+State+Act.
- Scope
- Benchmark with memory tracks
What it tests
- CECross-modal evidencePartially evaluated
- LELong-range evidenceDirectly evaluated
- SCState changesDirectly evaluated
- CFConflicting memoriesNot evaluated
- UQUnanswerable queriesNot evaluated
- MAMemory-grounded actionDirectly evaluated
- SMSpatial memoryDirectly evaluated
- OSObject-state trackingDirectly evaluated
- TSTask-state trackingPartially evaluated
Survey notes
The visual temporal-belief setting tests updating spatial information across observations.