The 48th issue of Zhiyuan LIVE will be held online from 11:00 to 12:00 on July 27, 2023. In this event, Sun Yutao, a doctoral candidate at Tsinghua University, will be invited to give an online report titled "RetNet Design Principles and Application Prospects" ".
Sun Yutao
Ph.D candidate at Tsinghua University
Sun Yutao will receive a bachelor's degree from Tsinghua University in 2023, and will continue to study for a doctorate in computer science in the same year, under the tutelage of Professor Wang Jianyong. At the same time, he has been conducting research work at Microsoft Research Asia since July 2022. His main research interests are the basic architecture of large models, modeling and reasoning of long texts, and the application of large models in other fields.
How does the chain of thought unlock and release the hidden ability of the big model
As a new neural network architecture, RetNet has strong modeling performance and inference speed, demonstrating its application potential as a natural language base. In this report, I will describe the design idea of RetNet, analyze the advantages and disadvantages of existing methods, and some conclusions in the experiment; in addition, the author will also introduce the plan to continue the work in the future, and in more scenarios down possibility.
Activity time: July 27 (Thursday) 11:00-12:00 (morning)
Activity form: online live broadcast, click "read the original text" to make an appointment; scan the QR code to enter the communication group
