Search papers, labs, and topics across Lattice.
1
0
3
You can slash DiT parameters by 70% and still beat larger models on text rendering, thanks to a smart compression framework that avoids training from scratch.
