LeRobot v0.6 vs v0.5: What Changed and How to Migrate
LeRobot v0.6 vs v0.5: world-model policies, GR00T N1.7, lerobot-rollout with DAgger, reward models, six benchmarks, and every breaking change.
LeRobotDataset v3.0: Chunked Parquet, MP4 Shards and Streaming for Robot Learning Data at Scale (2026)
LeRobotDataset v3.0 replaces per-episode files with chunked Parquet and MP4 shards plus streaming. What changes for robot-learning data pipelines at OXE scale.
Diffusion Policy for Robot Manipulation: Imitation Learning (2026)
Diffusion policy explained: how denoising diffusion models generate robot action sequences from demonstrations - visuomotor architecture, training, receding-horizon control and limits.