“A Retrieval-and-Structuring Approach for LLM-Enhanced, Theme-Focused Scientific Exploration” Abstract:  Large Language Models (LLMs) may bring unprecedented power for scientific exploration.  However, current LLMs may still encounter major challenges for effective scientific exploration due to their lack of in-depth, theme-focused data and knowledge.  Retrieval augmented generation (RAG) has recently become an interesting approach for augmenting LLMs with grounded, theme-specific datasets.  We discuss the challenges of RAG and propose a retrieval and structuring (RAS) approach, which enhances RAG by improving retrieval quality and mining structures (e.g., extracting entities and relations and building knowledge graphs) to ensure its effective integration of theme-specific data with LLM.  We show the promise of this approach at augmenting LLMs and discuss its potential power for LLM-enabled science exploration.  Bio: Jiawei Han is Michael Aiken Chair Professor in the Siebel School of Computing and Data Science, University of Illinois Urbana-Champaign.  He received ACM SIGKDD Innovation Award (2004), IEEE Computer Society Technical Achievement Award (2005), IEEE Computer Society W. Wallace McDowell Award (2009), Japan's Funai Achievement Award (2018), and being elevated to Fellow of Royal Society of Canada (2022).  He is Fellow of ACM and Fellow of IEEE and served as the Director of Information Network Academic Research Center (INARC) (2009-2016) supported by the Network Science-Collaborative Technology Alliance (NS-CTA) program of U.S. Army Research Lab and co-Director of KnowEnG, a Center of Excellence in Big Data Computing (2014-2019), funded by NIH Big Data to Knowledge (BD2K) Initiative.  Currently, he is serving on the executive committees of two NSF funded research centers:  MMLI (Molecular Make Research Institute)—one of NSF funded national AI centers since 2020 and I-Guide—The National Science Foundation (NSF) Institute for Geospatial Understanding through an Integrative Discovery Environment (I-GUIDE) since 2021.

Title: Transaction Processing in Hybrid Edge Data Management Systems Abstract: Advances in computing and storage systems have enabled end users to run complex workloads on relatively cheap machines. These advancements have given rise to a novel infrastructure in data management: Edge-Cloud. Edge-Cloud data management systems allow data to be stored at the edge and managed by […]

Abstract: The vast majority—over 80%—of data today exists in unstructured formats, and querying and extracting value from unstructured document collections remains a considerable challenge. While Large Language Models (LLMs) have made remarkable progress in document understanding, they fail to provide high-accuracy results for analytical queries on documents and incur high costs. In this talk, we […]