Retrieval And Structuring Augmented Generation with Large Language Models

🔍 Tutorial Overview: Retrieval-And-Structuring (RAS) Augmented Generation with LLMs

Large Language Models (LLMs) have transformed natural language processing, but their real-world application is often hindered by critical flaws like generating factually incorrect information (hallucinations), relying on outdated knowledge, and lacking expertise in specialized domains. 

This tutorial introduces Retrieval-And-Structuring (RAS) Augmented Generation, a cutting-edge paradigm designed to address these fundamental limitations. While standard Retrieval-Augmented Generation (RAG) grounds LLMs in external documents, it often struggles with unstructured text that can mislead the model. RAS evolves beyond RAG by integrating dynamic information retrieval with powerful knowledge structuring techniques. By converting unstructured text into organized representations like taxonomies and knowledge graphs, RAS creates a robust framework for LLMs to reason with greater accuracy, interpretability, and factual consistency. 

Join us for a comprehensive survey and hands-on exploration of the concepts, methods, and applications that are defining the next generation of knowledge-intensive AI. 

Our survey paper: paper link 

🔧 What You’ll Learn

• Foundations of RAG and RAS: Understand the evolution from naive RAG to advanced modular architectures. Learn how RAS addresses the core limitations of RAG by structuring retrieved information to guide structured LLM reasoning. 

• Information Retrieval Techniques

Dive into most advanced retrieval techniques: 

(1) Agentic Retrieval. Works covered: DeepRetrieval, Search-R1, s3

(2) Structure-Powered Intelligent Retrieval. Works covered: ToTER, TaxoIndex, Graph RAG, ..., KARE

• Text Structuring Methods: 

(1) Text Classification. Works covered: LOTClass, PIEClass, TELEClass

(2) Entity Typing and Entity Structure Mining. Works covered: OntoType, OneFET, RolePred, Reaction Miner, ActionIE, ZOES

(3) Relation Extraction and KG Construction. Works covered: TagReal, PriORE, GenRES, KG-FIT

• Structure-Augmented Reasoning/Generation: 

Works covered: StructRAG, KARE, RepoGraph, SARG, ClaimSpect, EpiMine

• Integration with LLMs: Explore how structured knowledge is integrated into LLM workflows using prompts, retrieval loops, and structure-aware reasoning frameworks.

• Applications and Case Studies: See how RAS powers breakthroughs in domains like healthcare, scientific discovery, and e-commerce.

• Open Challenges and Research Directions: Discover key research problems in retrieval efficiency, structure quality, and multi-modal, cross-lingual knowledge grounding.
  
  

🎯 Who Should Attend?

This tutorial is designed for researchers, practitioners, and advanced students working on:

• Information Retrieval

• Knowledge Graphs

• Agentic RAG.

• LLM-based applications

• Natural Language Processing

• Structure Mining

• Human-AI systems for knowledge-intensive tasks

⏰ Schedule

Part 1: An Introduction to LLM, RAG, and Its Application Exploration
1:00 – 1:20 PM

Part 2: Retrieving: Recent Advancements on Enhancing the Power of Retrieval [Slides]
1:20 – 2:15 PM

Part 3: Structuring: Knowledge Structuring to Help Retrieval and Augmented Generation [Slides]  
2:20 – 3:15 PM

Part 4: Reasoning: Enhancing LLM Augmented Generation with Structures
3:20 – 4:00 PM

Whether you’re building AI systems that need better knowledge access, or exploring structured methods to improve LLM performance, this tutorial will equip you with the concepts, tools, and recent advances to push the frontier.