Hugging Face updates LeRobot with world models and richer evaluation tools

Hugging Face released LeRobot v0.6.0, an update aimed at closing the loop between training, evaluation, deployment, and data collection in robot learning.

The release adds support for policies that can model future outcomes, reward models that judge task success, depth and language annotations for datasets, and new simulation benchmarks.

For robotics developers, the update makes LeRobot more useful as an end-to-end open stack rather than just a collection of models and examples.