Hugging Face released new LeRobot dataset quality metrics on October 8, 2026, saying it provides six key questions to help users evaluate robot training data before training a policy or fine-tuning a VLA. It is the company's first software update since the LeRobot dataset's initial release.
Hugging Face reported 10,000 episodes of robot training data, measured across multiple metrics, with 6,000 episodes pulled from the Hub. That compares with 400 episodes recorded by users in the past week.
LeRobot dataset quality metrics are built on metadata and Parquet files, targeting robot training data validation. Availability begins with the release of the metrics, initially for the robotics research community.
"You recorded 400 teleop episodes last week. You pulled 6,000 more from the Hub because the task looked close to yours. Now you want to fine-tune a VLA," said Harpreet Sahota, the author of the blog post.
The metrics help users find hidden issues in robot training data.
The announcement follows the release of the LeRobot dataset. Hugging Face said the metrics are a practical guide to improve dataset quality and avoid training errors.
Hugging Face did not say how the metrics will be integrated into existing workflows, and users are left to determine how to apply them. The company said the metrics will help improve dataset quality before training begins.
Source: huggingface
