1 paper
Hanxiao Wang, Biao Zhang, Jonathan Klein +3
We propose a transformer architecture and training strategy for tree generation. The architecture processes data at multiple resolutions and has an hourglass shape, with middle lay…