So101 Simple Cube Vary: Robot Telemetry for Task Learning

121 episodes of robot telemetry data generated using the LeRobot framework. The dataset contains 37,727 frames of video and associated data files, recorded at 30 frames per second. It was created by author Tacoin and last updated on June 17, 2026.

Use Cases

• Train robot control policies based on recorded visual and state data.
• Benchmark imitation learning algorithms using the provided episode demonstrations.
• Analyze robot performance on a single, consistent task across many episodes.

Strengths

• Contains 37,727 frames of video data across 121 episodes.
• Data is structured into 1000-frame chunks for easier processing.
• Includes both video files (200 MB) and structured data files (100 MB).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the structured data is unknown, which may limit suitability assessment.
• The dataset appears focused on a single task, limiting task diversity.

Provenance

Source

Generated using the LeRobot framework.

Collection Method

Simulated or recorded robot telemetry.

Freshness

Last updated 2026-06-17 19:19:48.