Please join us for a UW Data Science Seminar on Thursday, January 16th from 4:30 to 5:20 p.m. PT. The seminar will feature Xinyi Zhou, a postdoctoral scholar in the Paul G. Allen School of Computer Science and Engineering at the University of Washington.

The seminar will be held in the Physics/Astronomy Auditorium (PAA), Room A118 – campus map.

“Augmenting LLMs for Social Impact”

Abstract: Large language models (LLMs) are undeniably powerful, but often fall short in terms of trustworthiness and helpfulness in areas like factual accuracy and social intelligence. This can pose significant challenges when addressing complex societal issues such as health, democracy, science, and teamwork. In this talk, I will primarily introduce MUSE, an enhanced LLM that leverages web retrieval and multimodal integration to detect misinformation while delivering clear, accurate explanations backed by trustworthy references. MUSE outperforms top human benchmarks and leading LLMs like GPT-4. Additionally, I will present Social-RAG, a workflow that retrieves from group interactions to socially ground AI generation, which powers PaperPing—a deployed LLM agent that posts academic paper recommendations in group chats without disrupting existing social practices, fostering group common ground. I will conclude by discussing broader societal challenges and opportunities in the era of generative AI.

The 2024-2025 seminars will be held in person, and are free and open to the public.