Different Calculation Method
Research on calculation methods within large language models (LLMs) focuses on understanding how these models perform arithmetic and mathematical reasoning, aiming to improve their accuracy and efficiency. Current work investigates the internal mechanisms of LLMs, including the roles of attention heads and feed-forward networks, and explores the use of techniques like Fourier features and expected value calculations to enhance performance. These studies are significant because they shed light on the fundamental capabilities of LLMs and inform the development of more robust and reliable models for various applications, from improving reading comprehension to aiding medical image analysis.
Papers
September 21, 2024
September 6, 2024
June 11, 2024
June 5, 2024
April 25, 2024
April 22, 2024
March 9, 2024
September 6, 2023
August 22, 2023
August 2, 2023
June 10, 2023
May 24, 2023
May 8, 2023
October 24, 2022
August 31, 2022
May 24, 2022