Search papers, labs, and topics across Lattice.
1
0
3
5
Ditch the slow lane: $R^2$-dLLM turbocharges diffusion language models by slashing decoding steps by up to 75% without sacrificing quality.
