twanghcmutcc-by-4.0

WMBench — World-Model video benchmark for hallucination detection

Benchmark suite for evaluating hallucination and anomaly detection in videos generated by world models against real robot training distributions. Includes datasets from NVIDIA GR1 and planned Stanford DROID data with real and synthetically generated robot videos.

Downloads3K

Technical Profile

Modalities
rgbvideo
Robot Embodiments
NVIDIA GR1
Environment
lab
Task Types
manipulation
Data Format
mp4
License
cc-by-4.0
Part of the WMBench — World-Model video benchmark for hallucination detection family

Community Signals

Top 25% by downloads

Access

Need custom rgb data?

Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.

Request a Sample Pack

Related Datasets