VLA-RL-Study: What Can RL Bring to VLA Generalization? An Empirical Study

arXiv Website

This is the SFT model, fine-tuned from the warm-upped OpenVLA model. The SFT dataset consists of 16k trajectories collected by the motion planner. For more details, please refer to the codebase and the paper.

Downloads last month
8
Safetensors
Model size
8B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for gen-robot/openvla-7b-rlvla-sft_16k

Base model

openvla/openvla-7b
Finetuned
(4)
this model

Collection including gen-robot/openvla-7b-rlvla-sft_16k