Skip to content
arXiv cs.LG · Papers

LaPrune: Controllable Differentiable Sparsity at Million Scale

arXiv:2608.04057v1 Announce Type: new Abstract: Top-$k$ selection determines which components of a sparse model remain active. Hard selection blocks gradients, while continuous relaxations often couple mask hardness to the selected mass. We introduce LaPrune, a mathematically exact-budget differentiable layer that cont