Search papers, labs, and topics across Lattice.
1
0
2
Optimal block size selection can lead to a 4.20x speedup in diffusion-based speculative decoding, revolutionizing inference efficiency.
