Search papers, labs, and topics across Lattice.
1
0
3
2
Diffusion language models can achieve 400x faster inference without sacrificing quality by correcting self-conditioning errors that compound in few-step sampling.
