EfficientLLM: Scalable Pruning-Aware Pretraining for Architecture-Agnostic Edge Language Models Paper • 2502.06663 • Published Feb 10 • 1