1 paper
David Jansen, Roman Rausch, Ali Hashemi +2
In this paper, we formulate the compression of large language models (LLMs) by optimally deleting transformer blocks (``block removal'') as a constrained binary optimization (CBO)…