[2403.10799] Efficient Pruning of Large Language Model with Adaptive Estimation Fusion