PhD Candidate Yi (Joshua) Ren and Professor Danica Sutherland Recognized for Outstanding LLM Paper
UBC Computer Science PhD candidate Yi (Joshua) Ren and his supervisor, Professor Danica Sutherland, were recently recognized for an outstanding paper at the prestigious International Conference on Learning Representations (ICLR), held this year in Singapore. Their paper, titled "Learning Dynamics of LLM Finetuning," was one of only three selected as "Outstanding Papers" from 3,704 accepted submissions.
ICLR is one of the world's premier conferences in the field of deep learning and artificial intelligence. Outstanding papers are selected for their "theoretical insights, practical impact, exceptional writing, and experimental rigour."
Yi joined UBC after completing his master's degree at the University of Edinburgh. He now works with Professor Sutherland on cutting-edge research in machine learning and artificial intelligence. Their award-winning paper explores the "post-training" process of large language models (LLMs), the key phase that turns language predictors into (hopefully) useful chatbots and interactive agents. Their theoretical framework helps explain several surprising phenomena that can occur in this process and make the models behave poorly, setting the stage for more reliable methods in the future.
Congratulations to Yi and Danica on this significant achievement!
Further information: