🏢 Emory University

SparseLLM globally prunes large language models efficiently by decomposing the problem into manageable subproblems, achieving significant performance improvements, especially at high sparsity.