RAG Meets LLMs: Towards Retrieval-Augmented Large Language Models

About

Retrieval-augmented generation (RAG) has become a transformative approach for enhancing large language models (LLMs) by integrating external, reliable, and up-to-date knowledge. This addresses critical limitations such as hallucinations and outdated internal information. This tutorial delves into the evolution and frameworks of RAG, emphasizing the pivotal role of data management technologies in optimizing query processing, storage, indexing, and efficiency. It explores how RAG systems can deliver high-quality, context-aware outputs through efficient retrieval and integration, covering key topics such as retrievalaugmented LLM (RA-LLM) architectures, retrieval techniques, learning methodologies, and applications in NLP and domainspecific tasks. Challenges like customized query and generation, real-time retrieval, and trustworthy RAG are discussed alongside future directions and opportunities for innovation. Designed for students, researchers, and industry practitioners with basic artificial intelligence and data engineering knowledge, this tutorial offers practical insights into designing data management-powered RAG systems. It inspires the exploration of novel solutions in this rapidly evolving field.

Our Survey Paper: RAG-Meets-LLMs: Towards Retrieval-Augmented Large Language Models

Our slides: Part-1 and Part-2.

TARGET AUDIENCE AND PREREQUISITES FOR THE TUTORIAL

The audience of this tutorial could be college students, researchers in academic institutions, and industrial database and AI labs interested in data engineering, artificial intelligence, and the latest trends in LLMs. Audiences are expected to have some basic understanding of data engineering, artificial intelligence, information retrieval, and natural language processing. However, the tutorial will be presented at the college junior/senior level so that it can be comfortably followed by academic researchers or industrial practitioners who are interested in this emerging field. After attending this tutorial, the audience is expected to comprehensively understand data management in retrieval augmented generation and learn how to design data management-powered RAG for real-world applications. It is also hopeful that the audience gets inspired by this tutorial and can step further in this field to explore novel opportunities in data management-powered RAG.

Tutorial Syllabus

The topics of this tutorial include (but are not limited to) the following:

Retrieval-Augmented Generation (RAG)

Large Language Model (LLM)

Data Management

Pre-training

Fine-tuning

In-context Learning

Prompting

The tutorial outline is shown below:

Introduction of RA-LLMs (10 minutes)

Architecture of Combining RAG with LLMs (40 minutes)

RA-LLM architecture overview
Retriever in RA-LLMs
Retrieval results integration
Pre/Post-retrieval techniques
Special RA-LLM paradigms

Data Management for RAG (40 minutes)

Preliminaries of Data Management
RAG-powered by Data Management
Popular Vector Database Management Systems

Coffee Break (30 minutes)

Learning Approach of RA-LLMs (30 minutes)

Training-free methods
Training-based Methods

Applications of RA-LLMs (30 minutes)

NLP application
Downstream tasks
Domain-specific applications

Challenges and Future Directions of RA-LLMs (20 minutes)

Customized Query and Generation
Challenges in Real-Time Retrieval
Trustworthy RAG

Q&A (10 minutes)

About

TARGET AUDIENCE AND PREREQUISITES FOR THE TUTORIAL

Event Dates

Tutorial Syllabus

Organization

Tutorial TUTORS

Wenqi Fan

Pangjing Wu

Yujuan Ding

Shijie Wang

Liangbo Ning

Qing Li