A deeper look at depth pruning of LLMs