Search papers, labs, and topics across Lattice.
University of Michigan
2
0
4
0
BlazingAML delivers up to 333x speedups in anti-money laundering detection while maintaining state-of-the-art accuracy, by compiling high-level fuzzy pattern descriptions into optimized GPU code.
Masked diffusion models can finally achieve faster inference without sacrificing generation quality, thanks to a clever speculative decoding scheme.