Smolvla 2Cubes: Robot Teleoperation Data for Vision-Language-Action Models

A dataset created using the LeRobot framework for robot teleoperation. The meta information indicates a robot_type of 'so_follower' and includes video files totaling 200 MB, recorded at 30 frames per second. The dataset was authored by ydaichi and last updated on July 6, 2026.

Use Cases

• Train robot imitation learning models based on teleoperation data mentioned in the description.
• Benchmark vision-language-action model performance based on the 'so_follower' robot tasks.
• Develop robotic manipulation policies based on the multimodal video and state data structure.

Strengths

• Video data is recorded at 30 frames per second, which suggests smooth temporal resolution.
• The dataset structure is chunked into parquet files, which may facilitate efficient data loading.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

huggingface

Collection Method

Created using the LeRobot framework.

Freshness

Last updated 2026-07-06 10:08:12; freshness should be verified.