ghkim-rlwrldapache-2.0

shell_game — delta EEF, per-frame subtask

A robotic manipulation dataset where an arm lifts cups to find a hidden red block after shuffles, with per-frame subtask language annotations. Contains 45 episodes with 18,442 frames collected via teleoperation.

Downloads35
Episodes45

Why This Matters for Physical AI

This dataset demonstrates the importance of memory and temporal reasoning in robotic manipulation by requiring agents to track hidden object states across shuffled configurations that appear identical in the current frame.

Technical Profile

Modalities
rgbproprioception
Robot Embodiments
robotic arm with gripper
Action Space
end_effector_delta
Environment
lab
Task Types
manipulationcup liftingobject search
Episodes
45
Data Format
LeRobot
Annotation Types
language_instructionsaction_labelssubtask_labels
License
apache-2.0
Part of the shell_game 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