HighlightCurated by Aramai Editorial

HopRAG: Multi-Hop Reasoning for Logic-Aware Retrieval-Augmented Generation

A novel RAG framework that augments retrieval with logical reasoning through graph-structured knowledge exploration.

The paper proposes HopRAG, a framework that constructs a passage graph and employs a retriever-reason-prune mechanism to identify relevant passages based on logical connections. Experiments demonstrate improved final answer quality on multi-hop benchmarks. The framework expands the retrieval scope by exploring multi-hop neighbors guided by pseudo-queries and LLM reasoning.

Based on: HopRAG: Multi-Hop Reasoning for Logic-Aware Retrieval-Augmented Generation

HighlightCurated by Aramai Editorial

From knowledge graph construction to retrieval-augmented generation: a framework for comprehensive earthquake emergency support

A framework for comprehensive earthquake emergency support using large language models and knowledge representation techniques.

This study proposes a structured framework integrating LLMs with knowledge representation techniques to construct domain-specific knowledge graphs tailored for earthquake emergency scenarios. The framework extracts entities, relationships, and attributes from unstructured data and employs a knowledge fusion strategy to resolve ambiguities. A comparative experiment demonstrates the effectiveness of the Improved HybridRAG method in generating accurate and coherent responses.

Based on: From knowledge graph construction to retrieval-augmented generation: a framework for comprehensive earthquake emergency support · Geo-spatial Information Science

HighlightCurated by Aramai Editorial

MedRAG: Enhancing Retrieval-augmented Generation with Knowledge Graph-Elicited Reasoning for Healthcare Copilot

A retrieval-augmented generation model enhanced by knowledge graph-elicited reasoning for healthcare copilots.

This paper proposes MedRAG, a RAG model that integrates knowledge graphs and large language models to improve diagnostic accuracy in healthcare. It constructs a hierarchical diagnostic KG and retrieves EHRs to provide more accurate decision support. Experimental results show that MedRAG outperforms state-of-the-art models in reducing misdiagnosis rates.

Based on: MedRAG: Enhancing Retrieval-augmented Generation with Knowledge Graph-Elicited Reasoning for Healthcare Copilot

HighlightCurated by Aramai Editorial

Knowledge graph-extended retrieval augmented generation for question answering

A paper proposing a system that integrates Large Language Models and Knowledge Graphs for robust question answering.

The paper presents a system, called KG-RAG, which combines Large Language Models and Knowledge Graphs to improve question answering. It includes a question decomposition module and uses In-Context Learning and Chain-of-Thought prompting to generate explicit reasoning chains. Experiments show improved accuracy for multi-hop questions compared to baselines.

Based on: Knowledge graph-extended retrieval augmented generation for question answering · Applied Intelligence

HighlightCurated by Aramai Editorial

Personalizing Large Language Models using Retrieval Augmented Generation and Knowledge Graph

A paper proposing an approach to personalize large language models using retrieval augmented generation with knowledge graphs.

The authors propose a method to address over-fitting in large language models by incorporating knowledge graphs for personalized response generation. They use calendar data as an example of frequently updated personal information. Experimental results show improved accuracy and response time compared to baseline LLMs.

Based on: Personalizing Large Language Models using Retrieval Augmented Generation and Knowledge Graph

HighlightCurated by Aramai Editorial

Large Language Models for Structured and Semi-Structured Data, Recommender Systems and Knowledge Base Engineering: A Survey of Recent Techniques and Architectures

A survey of recent techniques and architectures using large language models.

This systematic review analyzes 88 studies on the use of large language models in recommendation systems, data processing, and knowledge base engineering. It highlights key trends and challenges, including hallucination mitigation, fairness, and domain adaptation. The review also considers the broader macroeconomic implications of deploying LLM-based systems.

Based on: Large Language Models for Structured and Semi-Structured Data, Recommender Systems and Knowledge Base Engineering: A Survey of Recent Techniques and Architectures · Electronics

HighlightCurated by Aramai Editorial

Fusion-Based Retrieval-Augmented Generation for Complex Question Answering with LLMs

A paper proposing a Retrieval-Augmented Generation model that integrates structured and unstructured knowledge.

The paper presents a dual-channel knowledge retrieval mechanism that targets structured and unstructured sources. A unified knowledge fusion network integrates both types of information into a coherent generation context, enhancing the accuracy and linguistic quality of generated outputs. The method shows strong stability and generalization in cross-domain tasks.

Based on: Fusion-Based Retrieval-Augmented Generation for Complex Question Answering with LLMs

HighlightCurated by Aramai Editorial

SkewRoute: Training-Free LLM Routing for Knowledge Graph Retrieval-Augmented Generation via Score Skewness of Retrieved Context

A training-free routing framework for knowledge graph retrieval-augmented generation.

The authors propose a simple and effective routing framework, SkewRoute, which balances performance and cost in knowledge graph retrieval-augmented generation. The framework is designed to direct queries to the most suitable language models based on score skewness of retrieved contexts. It achieves over 3x higher routing effectiveness while reducing runtime compared to existing methods.

Based on: SkewRoute: Training-Free LLM Routing for Knowledge Graph Retrieval-Augmented Generation via Score Skewness of Retrieved Context · arXiv (Cornell University)

HighlightCurated by Aramai Editorial

SkewRoute: Training-Free LLM Routing for Knowledge Graph Retrieval-Augmented Generation via Score Skewness of Retrieved Context

A training-free routing framework for knowledge graph retrieval-augmented generation.

The paper proposes a novel, training-free routing framework called SkewRoute. It is tailored to knowledge graph retrieval-augmented generation (KG-RAG) and effectively balances performance and cost. The method reduces calls to larger LLMs by up to 50% without sacrificing response quality.

Based on: SkewRoute: Training-Free LLM Routing for Knowledge Graph Retrieval-Augmented Generation via Score Skewness of Retrieved Context · Open MIND

HighlightCurated by Aramai Editorial

Knowledge Graph-Guided Retrieval Augmented Generation

Paper proposing a method for retrieval augmented generation using knowledge graphs.

This paper presents a method that combines knowledge graph-guided retrieval with augmented generation to improve the performance of language models.,The proposed approach uses a knowledge graph to guide the retrieval process and augment the generated text.,Experimental results demonstrate the effectiveness of the proposed method in improving the accuracy and coherence of generated text.

Based on: Knowledge Graph-Guided Retrieval Augmented Generation

HighlightCurated by Aramai Editorial

Transforming SHACL Shape Graphs into HTML Applications for Populating Knowledge Graphs

A paper proposing a method to design user interfaces for populating knowledge graphs using SHACL constraint files.

The authors present an approach to create multi-form web applications from SHACL constraints, enabling user interface modeling and leveraging OWL reasoning for logical consistency. This method treats editing knowledge graphs as a business process, integrating ontology-based components. The application models are themselves knowledge graphs that can be verified using OWL reasoning.

Based on: Transforming SHACL Shape Graphs into HTML Applications for Populating Knowledge Graphs · Digital

HighlightCurated by Aramai Editorial

FRAG: A Flexible Modular Framework for Retrieval-Augmented Generation based on Knowledge Graphs

A novel flexible modular framework for retrieval-augmented generation using knowledge graphs.

The paper proposes a flexible modular framework, FRAG, which improves retrieval quality while maintaining flexibility in knowledge graph-based retrieval-augmented generation. FRAG estimates the hop range of reasoning paths and applies tailored pipelines to ensure efficient and accurate reasoning path retrieval. The method does not require extra LLM fine-tuning or calls, significantly boosting efficiency and conserving resources.

Based on: FRAG: A Flexible Modular Framework for Retrieval-Augmented Generation based on Knowledge Graphs