Sheryl Hsu

Hsu studies how language models learn from feedback and use external information. She developed reinforcement learning from verbal feedback to preserve specific human corrections, and LeReT, which teaches models to improve retrieval by learning from the results of their own search attempts.