Article

Weekly Robotics #383: LoongForge-VLDA model, Robotheca open-source robot library, and Bedrock Robotics funding

Weekly Robotics issue #383 introduces LoongForge-VLDA, a vision-language-delta-action model built on LoongForge that extends π₀.₅, uses V-JEPA 2 visual representation learning to learn latent changes, and jointly trains action policies.

This page provides a summary and source link only.