← back to paper
arxiv: 2604.22050 · 2 revisions
LayerBoost: Layer-Aware Attention Reduction for Efficient LLMs