Feb 2, 2024 Friday
Abstract:
This talk will present how to use RL to enhance the abilities of LLM. We propose three types of method: the first is using rl style training process to fine-tune LLM by interacting with the env but without parameter updating. The second is use RL to gather external knowledge and combine it with llm’s internal knowledge to update the model. The third one is RLAIF, leverage self’s value to update the LLm.
Time:
Feb 2, 2024 Friday
11:00-11:50
Location:
Rm W1-101, GZ Campus
Online Zoom
Join Zoom at: https://hkust-gz-edu-cn.zoom.us/j/4236852791 OR 423 685 2791
Speaker Bio:
Dr. Lu Wang
Researcher, Microsoft Research Asia (MSRA)
Lu Wang is a researcher at Microsoft Research Asia (MSRA) specializing in llm, reinforcement learning, cloud intelligence, healthcare, finance, and other related areas. She received her PhD degree in East China Normal University (China) and has published over 20 papers on RL, including ICML, ICLR, and KDD.