Efficient Adaptation and Evaluation of LLMs

Speaker: Dr. Junxian HE
         Assistant Professor
         Department of Computer Science and Engineering Department
         The Hong Kong University of Science and Technology

Title:  "Efficient Adaptation and Evaluation of LLMs"

Date:   Monday, 27 November 2023

Time:   4:00pm - 5:00pm

Venue:  Lecture Theater F
        (Leung Yat Sing Lecture Theater)
        near lift 25/26, HKUST


Abstract:

As Large Language Models (LLMs) like ChatGPT gain traction in both
academic and industrial circles, addressing key factors to improve their
practicality is crucial. In this talk, I will delve into our latest
research endeavors focusing on two primary areas: efficient adaptation of
LLMs and their effective evaluation. Our discussion will encompass three
topics: (1) a training-free approach to compose off-the-shelf,
parameter-efficient modules, (2) a data profiling framework to achieve
data-efficient instruction tuning, and (3) C-Eval, the first comprehensive
Chinese benchmark intended to catalyze the development and growth of LLMs
for Chinese users.


****************
Biography:

Junxian He is an assistant professor at the Computer Science and
Engineering Department in the Hong Kong University of Science and
Technology. He received his PhD degree in natural language processing from
Carnegie Mellon University, Language Technologies Institute in 2022.
Before that, he received the bachelor degree in electronic engineering
from Shanghai Jiao Tong University in 2017. His recent research focuses on
efficient adaption, factuality, reasoning, and evaluation of large
language models. He has served as area chair for ACL and EMNLP, and his
work has been recognized by the Baidu PhD fellowship as well as the most
influential paper list in ICLR 2022.