prehjother

openarm v2 Confidence Labels

VLM-generated confidence labels indicating action chunk compressibility for bimanual robotic manipulation tasks (dustpan and scrub) using Gemini 3.8-Flash model.

Downloads23
Episodes107

Why This Matters for Physical AI

Provides VLM-based confidence labels for action quantization in bimanual manipulation, enabling training of adaptive compression policies for robotic task execution.

Technical Profile

Modalities
rgbproprioceptionlanguage
Robot Embodiments
bimanual_manipulator
Action Space
joint_positions
Environment
lab
Task Types
manipulationsweepingscrubbing
Episodes
107
Data Format
parquet
Annotation Types
language_instructionsaction_labelsconfidence_scores
License
other
Part of the openarm family

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