Search papers, labs, and topics across Lattice.
1
0
2
On-policy self-distillation can boost diffusion model performance by up to 44% while slashing training time by over 60%.
