Comparing RAG and GraphRAG for Page-Level Retrieval Question Answering on Math Textbook Comparing RAG and GraphRAG for Page-Level Retrieval Question Answering on Math TextbookTechnology-enhanced learning environments often help students retrieve relevant learning content for questions arising during self-paced study. Large language models (LLMs) have emerged as novel aids for information
Learning to Retrieve and Reason on Knowledge Graph through Active Self-Reflection Learning to Retrieve and Reason on Knowledge Graph through Active Self-ReflectionExtensive research has investigated the integration of large language models (LLMs) with knowledge graphs to enhance the reasoning process. However, understanding how models perform reasoning utilizing structured graph knowledge
In-depth Analysis of Graph-based RAG in a Unified Framework In-depth Analysis of Graph-based RAG in a Unified FrameworkGraph-based Retrieval-Augmented Generation (RAG) has proven effective in integrating external knowledge into large language models (LLMs), improving their factual accuracy, adaptability, interpretability, and trustworthiness. A number of graph-based RAG methods have been proposed
Why Ontologies are Key for Data Governance in the LLM Era? Reference Blog : https://medium.com/timbr-ai * 최근 LLM 도입을 검토하는 많은 기업 사이에서 온톨로지(Ontology)가 뜨거운 화두로 떠오르고 있습니다. 다들 잘 아시다시피, 온톨로지란 쉽게 말해 데이터 간의 관계와 의미를 정의하는 지도로 받아들일 수 있습니다. 단순히 데이터를 저장하는 것을
NodeRAG: Structuring Graph-based RAG with Heterogeneous Nodes NodeRAG: Structuring Graph-based RAG with Heterogeneous NodesRetrieval-augmented generation (RAG) empowers large language models to access external and private corpus, enabling factually consistent responses in specific domains. By exploiting the inherent structure of the corpus, graph-based RAG methods further enrich this process by building
When to use Graphs in RAG: A Comprehensive Analysis for Graph Retrieval-Augmented Generation When to use Graphs in RAG: A Comprehensive Analysis for Graph Retrieval-Augmented GenerationGraph retrieval-augmented generation (GraphRAG) has emerged as a powerful paradigm for enhancing large language models (LLMs) with external knowledge. It leverages graphs to model the
GraphFrames: Architectural Evolution from GraphX for Big Data and AI Applications GraphFrames: an integrated API for mixing graph and relational queriesGraph data is prevalent in many domains, but it has usually required specialized engines to analyze. This design is onerous for users and precludes optimization across complete workflows. We…OpenReview.
Verifying Chain-of-Thought Reasoning via Its Computational Graph Verifying Chain-of-Thought Reasoning via Its Computational GraphCurrent Chain-of-Thought (CoT) verification methods predict reasoning correctness based on outputs (black-box) or activations (gray-box), but offer limited insight into why a computation fails. We introduce a white-box method: Circuit-based Reasoning Verification (CRV). We hypothesize that attribution
Introducing GraphQA: An Agent for Asking Graphs Questions Introducing GraphQA: An Agent for Asking Graphs Questions | CatioGraphQA is Catio’s new open-source agent for natural-language questions over architecture graphs, fusing LLMs with graph algorithms to deliver fast, structure-aware answers for dependencies, flows, and system reasoning.catio-logo-blueIman MakaremiGitHub - catio-tech/graphqa:
GraphLand: Evaluating Graph Machine Learning Models on Diverse Industrial Data GraphLand: Evaluating Graph Machine Learning Models on Diverse Industrial DataAlthough data that can be naturally represented as graphs is widespread in real-world applications across diverse industries, popular graph ML benchmarks for node property prediction only cover a surprisingly narrow set
Flexible GraphRAG: a configurable open source framework for GraphRAG Blog link : https://integratedsemantics.org/ Github : https://github.com/stevereiner/flexible-graphrag * 이번 주 오마카세는 'Open Source Integrated AI and Semantic Tech' 블로그에서 설명하는 통합된 오픈소스 플랫폼 관련한 소식을 전달해드리고자 합니다. 최근 포스트에서는 문서 처리, 지식 그래프 구축, RAG 및
1.광고 안녕하세요 정이태입니다.오랜만에 10월 23일 목요일 저녁에 GUG 온라인 세미나가 예정되어 있습니다. 카이스트 김재철AI대학원 신기정 교수님 연구실 소속 김선우 연구원님과 하이퍼 그래프 이론부터 응용까지 라는 주제로 진행할 예정이며, 저또한 모더레이터 겸 요새 GraphRAG 트랜드부터 왜 HyperGraphRAG가 나오고 있는지 산업계 관점에서 생각을 공유드릴 예정이오니 관심있는 분들은 참여하셔서 본인 연구나
GraphOmakase
Topology of Reasoning: Understanding Large Reasoning Models through Reasoning Graph Properties Topology of Reasoning: Understanding Large Reasoning Models through Reasoning Graph PropertiesRecent large-scale reasoning models have achieved state-of-the-art performance on challenging mathematical benchmarks, yet the internal mechanisms underlying their success remain poorly understood. In this work, we introduce the notion
GraphOmakase
Are Large Language Models Good Temporal Graph Learners? Youtube link : https://www.youtube.com/watch?v=jmCwOQX9Ank * 이번 주 오마카세는 TGL (Temporal Graph Learning) 세미나에서 대규모 언어 모델(LLM)이 관계형 데이터베이스에서 어떻게 딥러닝을 수행하는지에 초점을 맞추어 발표된, 기존의 시계열 그래프 학습 방법론에 LLM을 접목하여 새로운 가능성을 탐색한 '대규모
GraphOmakase
GraphMETRO: Mitigating Complex Graph Distribution Shifts via Mixture of Aligned Experts GraphMETRO: Mitigating Complex Graph Distribution Shifts via Mixture of Aligned ExpertsGraph data are inherently complex and heterogeneous, leading to a high natural diversity of distributional shifts. However, it remains unclear how to build machine learning architectures that generalize to
GraphOmakase
Towards Quantifying Long-Range Interactions in Graph Machine Learning: a Large Graph Dataset and a Measurement Towards Quantifying Long-Range Interactions in Graph Machine Learning: a Large Graph Dataset and a MeasurementLong-range dependencies are critical for effective graph representation learning, yet most existing datasets focus on small graphs tailored to inductive tasks,
GraphOmakase
No Metric to Rule Them All: Toward Principled Evaluations of Graph-Learning Datasets No Metric to Rule Them All: Toward Principled Evaluations of Graph-Learning DatasetsBenchmark datasets have proved pivotal to the success of graph learning, and good benchmark datasets are crucial to guide the development of the field. Recent research has
GraphOmakase
Tensor-view Topological Graph Neural Networks paper link :https://arxiv.org/abs/2401.12007 official code : https://github.com/TaoWen0309/TTG-NN?utm_source=catalyzex.com * 저번 주 오마카세로 전해드렸던 텐서 연산의 그래픽 이해를 바탕으로, 이번 주 가볍게 소개해드릴 논문은 텐서 학습을 통합하여 로컬 및 글로벌 레벨에서 Tensor-view Topology 정보와 Tensor-view Graph의
GraphOmakase
Graph Tensor Networks: An Intuitive Framework for Designing Large-Scale Neural Learning Systems on Multiple Domain paper link : https://arxiv.org/abs/2303.13565 * 현재 토폴로지 신경망의 학습 메커니즘을 설계하는 과정에서 매우 중요한 텐서 연산에 대한 이해를 크게 도와준 논문 하나를 여러분들께 소개해드리려고 합니다. 그래프 구조를 활용하여 다양한 신경망의 텐서 연산
GraphOmakase
A General Graph Spectral Wavelet Convolution via Chebyshev Order Decomposition paper link : https://arxiv.org/abs/2405.13806 official code : https://github.com/liun-online/WaveGC * 그래프 스펙트럼 컨볼루션은 그래프 신호처리 이론을 기반으로 그래프 필터링, 데이터 분석 등의 넓은 분야에서 활용되고 있습니다. 다음은 그래프 스펙트럼 변환을 위한 신호 기저 (고유벡터) 선택
GraphOmakase
Mitigating Over-Squashing in Graph Neural Networks by Spectrum-Preserving Sparsification paper link : https://arxiv.org/abs/2506.16110 official code : https://github.com/Jinx-byebye/GOKU * ICML 2025 논문 리스트업 중 그래프 스펙트럼 관점에서 오버스쿼싱 문제를 해결하는 새로운 Graph Rewiring 논문을 가볍게 리뷰 전달을 해드리고자 합니다. 국내 그래프 연구에서 빼놓을 수 없는
GraphOmakase
Positional Encoding meets Persistent Homology on Graphs paper link : https://arxiv.org/abs/2506.05814 official code : https://github.com/Aalto-QuML/PIPE * 이번 주 오마카세에서는 그래프 표현 학습에서 중요한 모듈로써 활용되고 있는 '위치 인코딩(Positional Encoding, PE)'에 새로운 그래프 위상을 활용하는 점에서 제안된 논문을 소개해드리고자 합니다. * 이웃
GraphOmakase
Graph Representational Learning: When Does More Expressivity Hurt Generalization? paper link : https://arxiv.org/abs/2505.11298 official code : https://github.com/RPaolino/GenVsExp * 그래프 신경망(GNN)은 구조화된 데이터 학습에 매우 유용한 도구로 자리 잡았지만, GNN의 표현력(expressivity)과 실제 예측 성능(predictive performance) 사이의 관계는 여전히 명확하게 이해되지
GraphOmakase
On Measuring Long-Range Interactions in Graph Neural Networks paper link : https://arxiv.org/abs/2506.05971 official code : https://github.com/BenGutteridge/range-measure?utm_source=catalyzex.com * 이번주 오마카세에서는 오랜만에 고질적인 메세지 전달 메커니즘의 한계인 장거리 종속성(long-range dependency) 포착 문제를 해결하는 새로운 방법론을 제안한 논문을 소개해드리고자 합니다. * 다들 잘