Scale LLM
Scale in large language models (LLMs) is a critical research area focusing on improving efficiency, accessibility, and performance. Current efforts explore architectural innovations like Mixture-of-Experts (MoE) models to enhance speed and accuracy while managing computational costs, and methods for knowledge fusion to combine the strengths of multiple LLMs. These advancements aim to democratize access to powerful LLMs through open-source initiatives and efficient training strategies, ultimately impacting both scientific understanding of LLMs and their practical deployment in various applications.
Papers
October 24, 2024
September 14, 2024
July 18, 2024
May 23, 2024
February 25, 2024
January 23, 2024
October 30, 2023
September 15, 2023
August 2, 2023
July 17, 2023