LLM compression
Hierarchical Sparse Plus Low Rank: A New Approach to LLM Compress
New research introduces hierarchical sparse plus low rank compression for LLMs, combining structured sparsity with matrix decomposition for efficient model deployment.
arxiv research