We're looking to train better trf models on the updated data, yes. The work is still in progress though — plans might vary a bit.

→ View original post on X — @honnibal