Llama 3 Context-Obedient Models in a 3 x MoE configuration. The models experts are split into Understanding and summarizing input, following the format provided, and outputting only context relevant answers as the three experts involved 2ed3e35 verified TroyDoesAI commited on May 27