ResearchBenchmarksLanguage Modeling on Language Modeling Dataset PPL Llama-2-70BFollow3.32PerplexityDense3.25043.72024.194.6598May 23, 2025Evaluation ResultsMethodMethodLinksPerplexityDensePruning Ratio=0%Pruning Ratio=0%2025.053.32TRSP-ℓ2Pruning Ratio=25%Pruning Ratio=25%2025.054.13TRSP-ℓ1Pruning Ratio=25%Pruning Ratio=25%2025.054.28ShortGPTPruning Ratio=25%Pruning Ratio=25%2025.054.85LaCoPruning Ratio=25%Pruning Ratio=25%2025.054.92Shortened LLaMAPruning Ratio=25%Pruning Ratio=25%2025.054.98SLEBPruning Ratio=25%Pruning Ratio=25%2025.055.06