Do you have diagrams that contain more data points for those models?

[sictttsd]

Hi, I am trying to compare the performance of models focusing on predicting all joint state VS those predicting end effector poses and I saw the diagram list here. I believe OpenVLA and Octo are models predicting end effector poses (correct me if I am wrong), so my questions are

1. Do we have diagrams that have more data points for OpenVLA and Octo?
2. If I have a model that pretrained based on joint state data, what would happen if I finetune using sequences labeled with end effector poses, or vice versa?

Thanks for your time :)

[Giddy-oseko]

Yes i do