StructRAG: Boosting Knowledge Intensive Reasoning of LLMs via Inference-time Hybrid Information Structurization

2024/10/11 by Zhuoqun Li, Xuanang Chen, Li, Zhuoqun +17 · 1 voice · 29 citations
Computer Science · Decision Sciences · #Artificial Intelligence (cs.AI) #Artificial intelligence #Boosting (machine learning) #Computation and Language (cs.CL) #Computer science #Data Quality and Management #FOS: Computer and information sciences #Inference #Machine learning #Semantic Web and Ontologies #Topic Modeling #cs.AI #cs.CL

paper · pdf · doi:10.48550/arxiv.2410.08815

published in arXiv (Cornell University) (Cornell University)

openalex publication_date 2024/10/11 · arxiv published 2024/10/11 · openalex created_date 2024/10/16 · arxiv updated 2024/10/25 · openalex updated_date 2026/07/28

Abstract

Retrieval-augmented generation (RAG) is a key means to effectively enhance large language models (LLMs) in many knowledge-based tasks. However, existing RAG methods struggle with knowledge-intensive reasoning tasks, because useful information required to these tasks are badly scattered. This characteristic makes it difficult for existing RAG methods to accurately identify key information and perform global reasoning with such noisy augmentation. In this paper, motivated by the cognitive theories that humans convert raw information into various structured knowledge when tackling knowledge-intensive reasoning, we proposes a new framework, StructRAG, which can identify the optimal structure type for the task at hand, reconstruct original documents into this structured format, and infer answers based on the resulting structure. Extensive experiments across various knowledge-intensive tasks show that StructRAG achieves state-of-the-art performance, particularly excelling in challenging scenarios, demonstrating its potential as an effective solution for enhancing LLMs in complex real-world applications.

Cited by

• HyperbolicRAG: Enhancing Retrieval-Augmented Generation with Hyperbolic Representations
• Cog-RAG: Cognitive-Inspired Dual-Hypergraph with Theme Alignment Retrieval-Augmented Generation
• TabRAG: Improving Tabular Document Question Answering for Retrieval Augmented Generation via Structured Representations
• AGRAG: Advanced Graph-based Retrieval-Augmented Generation for LLMs
• Interact-RAG: Reason and Interact with the Corpus, Beyond Black-Box Retrieval
• AutoGraph-R1: End-to-End Reinforcement Learning for Knowledge Graph Construction
• Structure-R1: Dynamically Leveraging Structural Knowledge in LLM Reasoning through Reinforcement Learning
• PluriHop: Exhaustive, Recall-Sensitive QA over Distractor-Rich Corpora
• ReMindRAG: Low-Cost LLM-Guided Knowledge Graph Traversal for Efficient RAG
• G-reasoner: Foundation Models for Unified Reasoning over Graph-structured Knowledge
• Thinking Before You Speak: A Proactive Test-time Scaling Approach
• jXBW: A Compressed Index for Structure-Aware JSONL Retrieval in Structured RAG
• Structuring the Unstructured: A Systematic Review of Text-to-Structure Generation for Agentic AI with a Universal Evaluation Framework
• PaperRegister: Boosting Flexible-grained Paper Search via Hierarchical Register Indexing
• Towards Self-cognitive Exploration: Metacognitive Knowledge Graph Retrieval Augmented Generation
• LAG: Logic-Augmented Generation from a Cartesian Perspective
• SKA-Bench: A Fine-Grained Benchmark for Evaluating Structured Knowledge Understanding of LLMs
• Beyond Isolated Dots: Benchmarking Structured Table Construction as Deep Knowledge Extraction
• Towards Agentic RAG with Deep Reasoning: A Survey of RAG-Reasoning Systems in LLMs
• Structure-Augmented Reasoning Generation
• AdaVideoRAG: Omni-Contextual Adaptive Retrieval-Augmented Efficient Long Video Understanding
• When to use Graphs in RAG: A Comprehensive Analysis for Graph Retrieval-Augmented Generation
• Reason-Align-Respond: Aligning LLM Reasoning with Knowledge Graphs for KGQA
• From Single to Multi-Granularity: Toward Long-Term Memory Association and Selection of Conversational Agents
• Large Language models for Time Series Analysis: Techniques, Applications, and Challenges
• Finetune-RAG: Fine-Tuning Language Models to Resist Hallucination in Retrieval-Augmented Generation
• Disco-RAG: Discourse-Aware Retrieval-Augmented Generation
• Failure Modes in Multi-Hop QA: The Weakest Link Effect and the Recognition Bottleneck
• Credible Plan-Driven RAG Method for Multi-Hop Question Answering
• AgentAda: Skill-Adaptive Data Analytics for Tailored Insight Discovery

Discussions

• arxiv.org/abs/2410.08815 [bsky, 0 points, 0 comments]

Related

• RAS: Retrieval-And-Structuring for Knowledge-Intensive LLM Generation
• Struct-X: Enhancing Large Language Models Reasoning with Structured Data
• DeepRAG: Thinking to Retrieve Step by Step for Large Language Models
• A Survey on Retrieval And Structuring Augmented Generation with Large Language Models
• A Hybrid RAG System with Comprehensive Enhancement on Complex Reasoning