Search papers, labs, and topics across Lattice.
1
0
3
Masked diffusion models can finally achieve faster inference without sacrificing generation quality, thanks to a clever speculative decoding scheme.
