Reducing the Energy Dissipation of Large Language Models (LLMs) with Approximate Memories | Publicación