Search papers, labs, and topics across Lattice.
1
0
2
Achieving near-autoregressive accuracy while boosting decoding speed by over 2.4 times could redefine efficiency benchmarks in generative reasoning tasks.
