What Can RL Bring to VLA Generalization? An Empirical Study
Paper • 2505.19789 • Published • 1
This is the warm-upped model, fine-tuned from official openvla/openvla-7b.
The warm-up dataset consists of 140 trajectories collected by octo-small and the motion planner.
For more details, please refer to the codebase and the paper.
Base model
openvla/openvla-7b