RMBench
RMBench: Memory-Dependent Robotic Manipulation Benchmark with Insights into Policy Design
Tianxing Chen, Yuran Wang, Mingleyang Li, Yan Qin, Hao Shi, Zixuan Li, et al. · 2026-07-30
- Modality
- Embodied & 3D
- Data
- Simulated
- Task · ground truth
- Act. · State+Act.
- Scope
- Memory-specific benchmark
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 memoryPartially evaluated
- OSObject-state trackingDirectly evaluated
- TSTask-state trackingDirectly evaluated
Survey notes
Manipulation tasks explicitly vary the historical information needed beyond the current observation.