Know your robotics dataset before training.

Kalanos helps robotics data engineers inspect dataset quality, identify episodes that need review, and share assessments with their teams.

Try the LeRobot grader →   ·   Browse evaluated datasets →   ·   Get the CLI →

Explore an assessment

Start with ALOHA towel, PushT, or ALOHA insertion. Review the recorded findings, dataset revision, and assessment scope before comparing results.

From dataset to shared evidence

• Inspect: review episode-level outcomes and recorded data-quality findings.
• Curate: use those findings to decide which recordings need a closer look.
• Share: link to an assessment from your dataset card so others can inspect its evidence and limitations.

Working with private data? Run the CLI in your own environment and choose what to share.

Assessments describe the checks performed on a recorded revision. Unassessed capabilities are not passes, and a quality score does not guarantee training performance.

Website · Kalanos Hub · GitHub · Methodology