MarcelTornemit

setup_table_preferences

Real-robot trajectories for table setup on a Franka Panda with human pairwise preference labels across multiple judgment axes, designed for reward modeling and preference learning research.

Downloads0
Episodes467
Hours3.7

Why This Matters for Physical AI

This dataset enables preference learning and reward modeling research by providing structured human feedback on robot manipulation trajectories across multiple quality dimensions, advancing methods for learning from human preferences rather than scalar rewards.

Technical Profile

Modalities
rgbproprioception
Robot Embodiments
Franka Panda
Action Space
joint_positions
Environment
lab
Task Types
manipulationtable_setup
Episodes
467
Total Hours
3.7
Data Format
LeRobot
Annotation Types
pairwise_preferenceslanguage_instructionsreward_labels
License
mit
Part of the MarcelTorne preference learning datasets 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