VLA, Robot Foundation Models

OpenVLA

Resources

• https://openvla.github.io/
• https://arxiv.org/pdf/2406.09246
• https://github.com/openvla/openvla

They “train OpenVLA by fine-tuning a pretrained Prismatic-7B VLM” (Prismatic VLM), Prismatic follows the same standard architecture described above.

Uses:

• DINOv2
• SigLIP

They generate a 7D robot action:

• 3 degrees for position
• 3 for orientation
• 1 for gripper

Next

• OpenVLA-OFT