A proximal operator for a new 2:4 sparsity regularizer, plus masked gradient updates, improves post-training pruning of LLMs up to 70B.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Proximal Operator for Inducing 2:4-Sparsity
A proximal operator for a new 2:4 sparsity regularizer, plus masked gradient updates, improves post-training pruning of LLMs up to 70B.