Events
FDS Special Seminar
Scaling Human-Centric Trustworthy Foundation Model Reasoning
 |
Speaker: Yi R. (May) Fung (HKUST) Assistant Professor at the HKUST Department of Computer Science and Engineering Department Hong Kong University of Science and Technology Tuesday, November 19, 2024 4:00PM - 5:00PM Location: Yale Institute for Foundations of Data Science, Kline Tower 13th Floor, Room 1327, New Haven, CT 06511 and via Webcast: https://yale.hosted.panopto.com/Panopto/Pages/Viewer.aspx?id=e6783df6-2a96-4562-b898-b22901506dc9 |
Add To: Google Calendar | Outlook | iCal File
Abstract: In recent years, language models have made significant advancements, achieving remarkable performance on a large variety of tasks, as well as promising zero-shot/few-shot capabilities, bolstered by model scaling and innovative training techniques. Despite the exciting progress, ensuring these models align with fundamental constitutional principles of promoting helpful, honest, and harmless information communication remains a key challenge. In this talk, we present a roadmap for language models to function more responsibly and robustly. First, we introduce our method for enhancing model reasoning knowledge boundary awareness through instruction tuning that teaches large language models to say “I don’t know”. Then, we dive deeper into extending the enhancement of model knowledge boundary awareness from the language to multimedia domain, including augmenting the proactive engagement capabilities in model reasoning. We conclude this talk with a case study of language model self-play self-learning for greater personalization and scalable oversight, followed by a discussion of envisioned opportunities and exciting future research directions.
Speaker Bio: Yi R. (May) Fung is an Assistant Professor at the HKUST Department of Computer Science and Engineering Department, and holds an upcoming affiliate appointment at MIT. She completed her Ph.D. at the University of Illinois with Prof. Heng Ji. May’s research focus centers around socially-situated human-centered trustworthy NLP/AI with multimedia knowledge reasoning capability and scalable alignment principles. She is a recipient of three top AI/NLP conference paper awards, and is also organizing the AI4Research Workshop at AAAI’25 this year.
Hosted by Arman Cohan and Robert (Xiangru) Tang
Submit an Event
Interested in creating your own event, or have an event to share? Please fill the form if you’d like to send us an event you’d like to have added to the calendar.