CS/CLSP Seminar: Akari Asai

Description

Akari Asai, a doctoral candidate in the Paul G. Allen School of Computer Science & Engineering at the University of Washington, will give a talk titled "Beyond Scaling: Frontiers of Retrieval-Augmented Language Models" as a joint seminar for the Department of Computer Science and the Center for Language and Speech Processing.

Abstract:

Large language models (LLMs) have achieved remarkable progress by scaling training data and model sizes. However, they continue to face critical limitations, including hallucinations and outdated knowledge, which hinder their reliability—especially in expert domains such as scientific research and software development. In this talk, Akari Asai will argue that addressing these challenges requires moving beyond monolithic LLMs and toward augmented LLMs—a new artificial intelligence (AI) paradigm that designs, trains, and deploys LLMs alongside complementary modules to enhance reliability and efficiency. Focusing on her research on retrieval-augmented LLMs, one of the most impactful and widely adopted forms of augmented LLMs today, Asai will begin by presenting systematic analyses of current LLM shortcomings and demonstrating how retrieval augmentation offers a more scalable and effective path forward. She will then discuss her work on establishing new foundations for these systems, including novel training approaches and retrieval mechanisms that enable LLMs to dynamically adapt to diverse inputs. Finally, she will showcase the real-world impact of such models through OpenScholar, her group's fully open retrieval-augmented LLM for assisting scientists in synthesizing literature now used by over 30,000 researchers and practitioners worldwide. Asai will conclude by outlining her vision for the future of augmented LLMs, emphasizing advancements in abilities to handle heterogeneous modalities, more efficient and flexible integration with diverse components, and rigorous evaluation through interdisciplinary collaboration.