Curated Research & Literature
40 recent arXiv papers spanning all 6 GTI domains — surveyed, summarized, and mapped to the graph taxonomy we steward. Updated for 2024–2026.
Three Themes Shaping Graph Technology
LLM × Graph Convergence
Across nearly every domain, large language models are being integrated with graph technologies — from knowledge graph construction and Bayesian network discovery to causal graph inference and SPARQL query generation. This is the dominant research trend of 2025.
GNN Application Expansion
Graph Neural Networks continue to expand into new application domains: supply chains, IoT/6G, financial fraud, drug discovery, and protein interactions — moving well beyond their original social network and citation network roots.
Benchmark Maturation
Multiple comprehensive surveys and benchmark datasets have appeared for property graphs, community detection, AMR parsing, supply chains, and protein interactions — indicating these fields are standardizing, ideal for educational content.
Papers by Domain
Click a domain to filter, or browse all 40 papers below.
Data & Knowledge
9 papersLLM-empowered Knowledge Graph Construction: A Survey
Systematic analysis of how LLMs reshape the three-layered KG construction pipeline — ontology engineering, knowledge extraction, and knowledge fusion.
A Survey of Graph Retrieval-Augmented Generation for Customized LLMs
Introduces GraphRAG as a paradigm overcoming traditional RAG limitations through graph-structured knowledge representation, enabling multihop reasoning.
A Survey on Spatio-Temporal Knowledge Graph Models
Surveys STKGs that integrate entities, relationships, time, and space for environmental systems, urban planning, and transportation.
Survey: Graph Databases
Comprehensive survey examining property graph models, query languages (Cypher, PGQL, GQL), and storage architectures.
Multi-Agent GraphRAG: Text-to-Cypher for Labeled Property Graphs
Multi-agent framework translating natural language questions into Cypher queries over labeled property graphs.
Learning-Augmented Online Bipartite Matching
Advances algorithmic theory for bipartite graph optimization with ML-augmented decision-making in the random arrival model.
Understanding Survey Paper Taxonomy via Graph Representation Learning
Uses graph representation learning on co-category structures to classify papers within a taxonomy, outperforming language models.
TELEClass: Taxonomy Enrichment and LLM-Enhanced Hierarchical Text Classification
Combines LLMs with taxonomy enrichment for hierarchical text classification using graph-structured label hierarchies.
SPARQL Query Generation with LLMs: Training Data Memorization & Knowledge Injection
Evaluates LLM-generated SPARQL queries for question answering over knowledge graphs and linked data endpoints.
AI & Machine Learning
8 papersGNNs in the Real World: Imbalance, Noise, Privacy and OOD Challenges
Reviews GNN solutions to four critical real-world challenges: class imbalance, noisy data, privacy constraints, and out-of-distribution generalization.
Graph Neural Networks for Next-Generation IoT
Explores GNN deployment in next-gen IoT systems with 6G technologies — massive MIMO, RIS, THz communication, and satellite systems.
Optimizing Automatic Differentiation with Deep Reinforcement Learning
Frames automatic differentiation as an ordered vertex elimination problem on the computational graph, using deep RL for efficient elimination.
Enabling Automatic Differentiation with Mollified Graph Neural Operators
First method to leverage automatic differentiation and compute exact gradients on arbitrary geometries via graph neural operators.
Bayesian Network Software Packages: Structure & Parameter Learning (2025 Ed.)
Reviews the most relevant tools for Bayesian Network structure and parameter learning with practical recommendations.
Bayesian Network Structure Discovery Using Large Language Models
Unified framework (PromptBN and ReActBN) placing LLMs at the center of Bayesian network structure discovery.
Survey of Abstract Meaning Representation: Then, Now, Future
Comprehensive survey of AMR parsing (text-to-AMR) and generation (AMR-to-text), with applications in NLP tasks.
Finetuned LLMs in Abstract Meaning Representation Parsing
Demonstrates that finetuning decoder-only LLMs achieves comparable performance to complex AMR parsers, simplifying graph construction.
Social & Network
7 papersA Comprehensive Review of Community Detection in Graphs
Thorough exposition of community detection methods: modularity-based, spectral clustering, probabilistic modeling, and deep learning.
Community Detection Robustness of Graph Neural Networks
Evaluates robustness of GNN-based community detection under adversarial perturbations and noisy graph structures.
Graph Neural Networks for Financial Fraud Detection: A Review
Unified framework categorizing GNN methodologies for financial fraud detection, significantly outperforming traditional methods.
ATM-GAD: Adaptive Temporal Motif Graph Anomaly Detection
Temporal motif extraction and dual-attention blocks detect multi-step fraud schemes in financial networks.
DynBERG: Dynamic BERT-based GNN for Financial Fraud Detection
Integrates Graph-BERT with GRU layers to capture temporal evolution across subgraphs in directed financial transaction networks.
Structure-prior Informed Diffusion Model for Graph Source Localization
Generative diffusion framework with topology-aware priors for robust source localization in information propagation networks.
Diffusion Model Agnostic Social Influence Maximization in Hyperbolic Space
Influence maximization methods using hyperbolic space representations that capture hierarchical social influence structures.
Infrastructure
5 papersGATES: Cost-aware Dynamic Workflow Scheduling via GATs and Evolution Strategy
Cloud workflow scheduling where DAG tasks are assigned to VMs using deep RL with graph attention networks learning topological relationships.
Memory-aware Adaptive Scheduling of Scientific Workflows on Heterogeneous Architectures
Scientific workflows as DAGs achieving optimal makespan under memory constraints on heterogeneous processors.
DAG-Plan: Generating DAGs for Dual-Arm Cooperative Planning
Automatically generates directed acyclic dependency graphs for robotic task planning, enabling cooperative execution.
GNNs in Supply Chain Analytics: Concepts, Dataset and Benchmarks
Establishes supply chains as inherently graph-like, presents a multi-perspective real-world FMCG benchmark dataset.
Graph-based Digital Twins for Supply Chain Management and Optimization
Combines graph modeling with Digital Twin architecture for dynamic, real-time supply network representations.
Science & Research
7 papersCausal MAS: LLM Architectures for Discovery and Effect Estimation
Surveys multi-agent systems leveraging LLM agents for causal reasoning in scientific discovery, healthcare, and fact-checking.
Large Language Models for Causal Discovery: Current Landscape & Future
Examines how LLMs transform causal discovery: direct extraction, domain knowledge integration, and structure refinement.
Causal Graphs with Latent Confounders from Interventional Data
Advances interventional causal discovery by identifying causal relations through distributional changes, even with latent confounders.
GNNs for Drug Discovery: Recent Developments and Challenges
Comprehensive survey spanning molecular property prediction, virtual screening, molecular generation, and synthesis planning.
Graph-structured Small Molecule Drug Discovery Through Deep Learning
Reviews deep learning on graph-structured molecular representations for property prediction, de novo generation, and reaction prediction.
PRING: Rethinking Protein-Protein Interaction Prediction from Pairs to Graphs
First comprehensive benchmark evaluating PPI prediction from a graph-level perspective (21,484 proteins, 186,818 interactions).
Extracting Inter-Protein Interactions Via Multitasking Graph Structure Learning
Graph attention mining of protein structural information with multitask learning to advance PPI prediction accuracy.
Semantic Web
4 papersOntology Learning and Knowledge Graph Construction: A Comparison
Compares approaches to automated ontology learning, demonstrating how LLMs enable ontology extraction from structured and unstructured data.
Ontology-Enhanced Knowledge Graph Completion using LLMs
Demonstrates how ontological information in RDF triples enhances knowledge graph completion through LLMs.
Automating SPARQL Query Translations between DBpedia and Wikidata
Evaluates LLM performance on SPARQL-to-SPARQL translation between major linked data knowledge graphs.
The KnowWhereGraph Ontology
Large-scale geospatial knowledge graph built on RDF and linked data principles for environmental and humanitarian applications.