The Emergence of Essential Sparsity in Large Pre-trained Models: The Weights that Matter Anonymous Author(s) Affiliation Address email 1 Supplementary Material 1

Accuracy F1-score Accuracy (Top-1) 1.2 SMC-Bench Arithmetic reasoning T ask Settings Table 2: Hyperparameters and training configurations used for models on Arithmetic Reasoning.Datasets MA VPS, ASDiv-A, SV AMP Pre-trained Embeddings bert-base Embedding Size [768] Hidden Size [384] Number of Layers...