Keynote Speakers

We are very excited to have Zhuyun Dai and Qi Chen as keynote speakers for ReNeuIR 2024! Below you will a short bio for each speaker; titles and abstracts will be made available soon.

Zhuyun Dai

LLM-Powered Retrieval: From Distillation to New Architectures

Abstract

Information retrieval systems are essential for accessing the vast knowledge stored in large corpora, but current models often fall short when it comes to reasoning, following instructions, and generalizing to new distributions. This talk delves into our research aimed at enhancing retrieval models by harnessing the power of large language models (LLMs). We first tackle the challenge of generalizing neural retrievers across different domains. We showcase how LLM distillation can be leveraged to achieve this, enabling versatile neural retrievers like the Gecko text embeddings API. Next, we introduce XTR, a novel multi-vector retrieval approach that brings closer the architectures of LLMs and retrievers. XTR significantly improved efficiency compared to previous token-level retrieval methods. Finally, we explore the potential of long-context LLMs to revolutionize the future of retrieval by digesting the entire corpus as a prompt. To evaluate this exciting frontier, we introduce LOFT, a new benchmark specifically designed to assess the impact of long-context models retrieval, retrieval-augmented generation (RAG), and database querying.

Bio

Zhuyun Dai is a Staff Research Scientist at Google DeepMind. Her research interests lie in large language models, information retrieval, and machine learning. Recently, her work has concentrated on developing generalizable and capable neural retrieval models, enhancing large language model factuality, and advancing instruction fine-tuning techniques. Additionally, Zhuyun actively contributes to the organizational community of various conferences in these fields.

Zhuyun earned her Ph.D. from the Language Technologies Institute at Carnegie Mellon University in 2020, under the supervision of Prof. Jamie Callan. She holds an undergraduate degree in Computer Science from Peking University.

Zhuyun

Qi Chen

Rethink Information Retrieval in the AI Era

Abstract

Although large language models (LLMs), a breakthrough in the field of artificial intelligence, have provided a novel way to learn and access an extensive scale of knowledge, they still exhibit certain limitations, such as hallucination and absence of fresh information. Therefore, integrating an external up-to-date information-rich knowledge database with large language models is of paramount importance to enhance their performance and reliability. We observe that combining scalar queries with vector queries enables much more powerful semantic-rich analytics that was previously difficult if not impossible, which requires unification of vector database and conventional database. In this talk, we introduce SPANN, SPFresh, VBase and OneSparse to address the index scalability, freshness, and unification foundation these three long-standing problems, which unleash the AI power with fresh and accurate information retrieval support. In addition, we introduce MS MARCO Web Search, the first large-scale information-rich web dataset, featuring millions of real clicked query-document labels. We hope MS MARCO Web Search can serve as a benchmark for modern web-scale information retrieval, facilitating future research and innovation in diverse directions.

Bio

Qi Chen is currently a Principal Researcher of Microsoft Research Asia Vancouver. She obtained her Bachelor’s and Ph.D. degrees from the School of Computer Science at Peking University, under the guidance of Professor Xiao Zhen. From 2013 to 2014, she worked as a visiting student under the supervision of Professor Jinyang Li at the System Group of New York University. She is a recipient of the OSDI’20 Best Paper Award and the NeurIPS’22 Outstanding Paper Award. Her current research interests include distributed systems, cloud computing, vector database and deep learning algorithms and artificial intelligence systems.