Search papers, labs, and topics across Lattice.
1
0
3
Train smarter, not harder: DSL unlocks 4x faster non-autoregressive generation by teaching masked diffusion models to self-correct more efficiently.
