← Home

Daily arXiv Papers

Graph Neural Networks · Graph Learning · LLM × Graph
Conferences
Archives
2026-08-11 2026-08-10 2026-08-09 2026-08-08 2026-08-07 2026-08-06 2026-08-05 2026-08-04 2026-08-03 2026-08-02 2026-08-01 2026-07-31 2026-07-30 2026-07-29 2026-07-28 2026-07-27 2026-07-26 2026-07-25 2026-07-24 2026-07-23 2026-07-22 2026-07-21 2026-07-20 2026-07-19 2026-07-18 2026-07-17 2026-07-16 2026-07-15 2026-07-14 2026-07-13 2026-07-12 2026-07-11 2026-07-10 2026-07-09 2026-07-08 2026-07-07 2026-07-06 2026-07-05 2026-07-04 2026-07-03 2026-07-02 2026-07-01 2026-06-30 2026-06-29 2026-06-28 2026-06-27 2026-06-26 2026-06-25 2026-06-24 2026-06-23 2026-06-22 2026-06-21 2026-06-20 2026-06-19 2026-06-18 2026-06-17 2026-06-16 2026-06-15 2026-06-14 2026-06-13 2026-06-12 2026-06-11 2026-06-10 2026-06-09 2026-06-08 2026-06-07 2026-06-06 2026-06-05 2026-06-04 2026-06-03 2026-06-02 2026-06-01 2026-05-31 2026-05-30 2026-05-29 2026-05-28 2026-05-27 2026-05-26 2026-05-25 2026-05-24 2026-05-23 2026-05-22 2026-05-21 2026-05-20 2026-05-19 2026-05-18 2026-05-17 2026-05-16 2026-05-15 2026-05-14 2026-05-13 2026-05-12 2026-05-11 2026-05-10 2026-05-09 2026-05-08 2026-05-07 2026-05-06 2026-05-05 2026-05-04 2026-05-03 2026-05-02 2026-05-01 2026-04-30 2026-04-29 2026-04-28 2026-04-27 2026-04-26 2026-04-25 2026-04-24 2026-04-23 2026-04-22 2026-04-21 2026-04-20 2026-04-19 2026-04-18 2026-04-17 2026-04-16 2026-04-15 2026-04-14 2026-04-13 2026-04-12 2026-04-11 2026-04-10 2026-04-09 2026-04-08 2026-04-07 2026-04-06 2026-04-05 2026-04-04 2026-04-03 2026-04-02 2026-04-01 2026-03-31 2026-03-30 2026-03-29 2026-03-28 2026-03-27 2026-03-26 2026-03-25 2026-03-24 2026-03-23 2026-03-22 2026-03-21 2026-03-20 2026-03-19 2026-03-18 2026-03-17 2026-03-16 2026-03-15 2026-03-14 2026-03-13 2026-03-12 2026-03-11 2026-03-10 2026-03-09 2026-03-07 2026-03-06 2026-03-05 2026-03-04 2026-03-03 2026-03-02 2026-03-01 2026-02-28 2026-02-27 2026-02-26 2026-02-25 2026-02-24 2026-02-23 2026-02-22 2026-02-21 2026-02-20 2026-02-19 2026-02-18 2026-02-17 2026-02-16 2026-02-15 2026-02-14 2026-02-13 2026-02-12

Showing 32 papers for 2026-08-05

PatTree: a novel approach for automated creation of multimodal, graph-based patient representations for medical classification tasks
Graph Learning

PatTree proposes an automated method to build multimodal, graph-based patient representations for medical classification. It aims to tackle real-world data challenges—heterogeneity, missing values, multiple time points, and diverse modalities—through data harmonization and integrated graph representations.

GoT-CD: Graph-of-Thoughts Causal Discovery and the Fragility of Post-hoc Path-Specific Fairness Audits
Graph Learning Graph Theory

GoT-CD introduces Graph-of-Thoughts for causal discovery and highlights how post-hoc path-specific fairness audits are fragile because the estimands depend on a supplied causal graph. Errors in the discovery step propagate to fairness judgments. It also critiques how current evaluation metrics weight edges equally, potentially masking edge-level issues.

When Should Graph Attention Be Sparse? Learning a Per-Edge Tsallis Index
GNN Graph Learning

LTGA proposes a graph attention layer with a learnable Tsallis entropy index q for each edge, allowing attention to interpolate between heavy-tailed, softmax, and compact-support shapes. This per-edge adaptability helps the model tailor attention to both homophilic and heterophilic graphs.

A Graph Signal Processing Perspective on Numerical Sequence Representations in LLM In-Context Learning
Graph Theory Graph Learning

From a graph signal processing perspective, the paper studies numerical sequence representations in in-context learning, treating attention as a graph over tokens. It analyzes how numerical information is organized within LLM representations beyond output-level evaluations.

Learning and Clustering on Temporal Graphs: Principles, Primitives, and Pooling
Graph Learning

Focuses on learning and clustering on temporal graphs, introducing principles, primitives, and pooling operators to obtain coarse-grained representations that capture temporal dynamics. It compares these methods against traditional clustering and standard GNN approaches to assess benefits.

Foundations of Equivariant Deep Learning: Unifying Graph and Sheaf Neural Networks
GNN Graph Learning

Foundations of Equivariant Deep Learning introduces order-equivariant neural networks (OENN) that unify graph message passing with sheaf neural networks under the theory of equivariant bundles. This framework broadens geometric deep learning to incorporate richer symmetry structures beyond simple group actions.

The Graph Language: How Knowledge Graphs Speak to Large Language Models
Knowledge Graph LLM × Graph

GRALAN enables knowledge graphs to communicate with LLMs by introducing relational tokens that preserve graph structure in the LLM's semantic space. A trainable language mediator converts KGs into tokens suitable for frozen LLMs, enabling grounded, graph-aware reasoning. Demonstrated in QA by reframing tasks as entity classification.

G-ReAct: Graph-Guided Deep Search via Structure-State Co-Evolution
Graph Learning

G-ReAct proposes graph-guided deep search with structure-state co-evolution to address forgetting, drift, and exploration inefficiency in long-horizon reasoning.

CockpitHAT: Dependency-Graph-Driven Hierarchical Attribution for Embodied Multi-Agent Cockpits
Graph Learning

CockpitHAT introduces hierarchical attribution for embodied multi-agent systems in automotive cockpits, using dependency graphs and multi-channel evidence to overcome limitations of text-trace-based attribution. It emphasizes safety-aware evaluation.

CoEvoKG: Co-Evolving Knowledge Graphs with Self-Evolving Search Agents
Knowledge Graph Graph Learning

Introduces CoEvoKG, a framework where a knowledge graph serves both as a source of verifiable training tasks and a persistent evidence memory for evolving search agents. It jointly trains a task generator and a search agent: the generator creates multihop questions from KG entity chains; the agent learns from these tasks.

HPFA: Hypergraph-Based Paired Failure Attribution for LLM Reasoning
Graph Learning

HPFA introduces hypergraph-based paired failure attribution to attribute failures to specific reasoning steps by exploiting nonlinear dependencies in reasoning traces; more scalable than stepwise counterfactual tests.

SkillTrace: Traversing a Query-Skill Graph for Composable LLM Agents
Graph Theory

SkillTrace organizes a three-level query-skill graph to identify complete executable skill compositions for composable LLM agents.

KING: Embodiment-Aware Kinematic Graph Neural Network for Unified Motion Representation of Legged and Wheeled Robots
GNN Graph Learning

KING proposes a kinodynamic GNN that provides a unified motion representation across legged and wheeled robots, leveraging kinematic constraints for robust odometry in featureless environments. It targets better generalization across embodiments by accounting for varied proprioception.

SG-Layout: Structured Scene Graph-Guided Layout Generation with LLMs
Graph Theory

SG-Layout introduces structured scene graph-guided layout generation for LLMs, addressing the difficulty of encoding explicit geometric relationships. It uses a two-stage training: graph-language feature alignment to align relational graph with language features.

X-KGRank: A Knowledge Graph RAG Framework for Explainable Recommendations via Pattern Mining and LLM Re-Ranking
Knowledge Graph LLM × Graph

X-KGRank is a knowledge graph retrieval augmented framework for explainable recommendations; mining patterns in KG and re-ranking with LLMs; demonstrates with MovieLens-1M.

Multi-Source Dynamic Graph Learning for Compound-Flood Forecasting in Managed Coastal Systems
Graph Learning

Multi-Source Dynamic Graph Learning for Compound-Flood Forecasting uses signals from multiple monitoring stations; models interactions with a dynamic graph; aims to forecast prolonged high-water plateaus.

LAB-Tab: LLM-Augmented Bayesian Network Adaptation for Few-Shot Tabular Generation
Graph Theory

LAB-Tab merges LLMs with Bayesian network adaptation for few-shot tabular generation; balances target-domain fidelity and diversity; leverages target and source domain information.

Effective and Efficient Context Retrieval via Partial Dependency Graph for Repository-Level Code Generation
Graph Learning

Effective and Efficient Context Retrieval via Partial Dependency Graph for Repository-Level Code Generation uses partial dependency graphs to retrieve context, improving coverage of dependencies beyond simple similarity.

TBSG-Net: Temporal Bipartite Scene Graph Network for Fine-Grained Video Moment Retrieval
GNN Graph Learning

TBSG-Net introduces Temporal Bipartite Scene Graph Network to model temporal evolution of objects and relations for fine-grained video moment retrieval; addresses lack of temporal dynamics and explicit temporal spans.

Knowledge Graph Augmented Large Language Models for Disease Prediction
Knowledge Graph LLM × Graph

Knowledge Graph Augmented LLMs for Disease Prediction uses KG-guided chain-of-thought for visit-level disease prediction on MIMIC-III, maps ICD-9 to PrimeKG, mines relevant nodes/paths, scaffolds temporally consistent CoT rationales, and fine-tunes lightweight LLMs.

Panning for Gold: Expanding Domain-Specific Knowledge Graphs with General Knowledge
Knowledge Graph

Domain-specific knowledge graphs (DKGs) are critical yet often suffer from limited coverage compared to General Knowledge Graphs (GKGs). Existing enrichment tasks rely on external unstructured data or internal reasoning, but integration scope and quality remain limited. This paper argues the gap and proposes systematic ways to leverage high-quality GKGs to supplement DKGs.

OntoTKGE: Ontology-Enhanced Temporal Knowledge Graph Extrapolation
Knowledge Graph

We propose OntoTKGE, an ontology-enhanced temporal KG extrapolation model. The key idea is to alleviate sparsity by enabling entities to inherit behavioral patterns from others that share the same concept via ontological knowledge. The approach uses a novel encoder–decoder framework to fuse temporal data with ontological priors.

Unifying biomedical knowledge in a modern multimodal graph
Knowledge Graph

OptimusKG is a multimodal biomedical labeled property graph built from structured and semi-structured resources to preserve factual, type-specific metadata across molecular, anatomical, clinical, and environmental domains. It unifies disparate sources into a coherent schema and contains about 190,939 nodes with rich metadata.

The Violation Situation Pattern: Persistent Representation of Compliance Violations in Knowledge Graphs
Knowledge Graph

We introduce the Violation Situation Pattern (VSP), a reusable ontology pattern that treats each detected compliance violation as a persistent first-class graph entity. Each violation carries a rule identifier, a temporal validity interval, a lifecycle state, and evidence links to affected processes.

Fairness in Augmented Graph Learning: A Survey
Graph Learning

This paper surveys fairness in augmented graph learning (AGL), which integrates techniques like federated learning, graph transformers, and graph condensation. It introduces FairGX and discusses how traditional GNN debiasing methods fall short in AGL settings, outlining key fairness challenges and potential mitigations.

Representing Higher-Order Networks: A Survey of Graph-Based Frameworks
Graph Theory

This book surveys higher-order network formalisms that extend classical pairwise graphs to capture richer interactions, including multiway, hierarchical, temporal, multilayer, recursive, and tensor-based frameworks. The goal is to enable more expressive representations of complex systems.

LongCrafter: Towards Diverse Long-Context Understanding via Evidence-Graph-Guided Instruction Synthesis
LLM × Graph Graph Learning

LongCrafter presents a framework for generating diverse long-context supervised fine-tuning data to improve long-context understanding in LLMs. It couples a hierarchical task taxonomy with an evidence-grounded pipeline to ensure task coverage, difficulty, and faithfulness supervision.

Dimensionality Reduction Meets Network Science: Sensemaking on UMAP's kNN Graph
Graph Theory

We explore the internal kNN graph constructed by UMAP, showing that standard graph algorithms on this graph (beyond the final 2D embedding) can enhance sensemaking. Examples include using PageRank to identify representative data points and other analyses that reveal data topology prior to projection.

Fault Cause Identification across Manufacturing Lines through Ontology-Guided and Process-Aware FMEA Graph Learning with LLMs
Graph Learning

We address fault cause identification in complex manufacturing lines using ontology-guided and process-aware FMEA graph learning with LLMs. The FMEA worksheets contain valuable insights but are hard to reuse due to language variability and process differences. The proposed approach uses ontologies and process-aware graph learning with LLMs to improve cross-configuration fault diagnosis.

Fast and Efficient Approximate Nearest Neighbor Search for High-Dimensional LLM Embeddings
Graph Theory

We submit for SISAP 2026 on fast and efficient approximate nearest neighbor search for high-dimensional LLM embeddings. Task 1: construct kNN graphs for 1024-d BGE-M3 embeddings; Task 2: perform MIPS on unnormalized Llama-3.2-8B features. We use Equi-Voronoi Polytopes (EVP) for fast quantization, aided by reranking to preserve accuracy.

Extracting ODRL Policies from Business Process Models: A Graph Traversal Approach to Compliance-by-Extraction
Graph Theory

We propose a graph traversal approach to extract ODRL policies from BPMN process models, enabling Compliance-by-Extraction. BPMN models encode normative content, and ODRL is a machine-readable policy language; the approach automatically derives ODRL policies from process graphs to support scalable compliance.

PULSE: An Executable Contract Language for Spatiotemporal Knowledge Graph Engineering
Knowledge Graph

We present PULSE, an executable contract language for spatiotemporal KG engineering, inspired by Object-Process Methodology. The language localizes four operational roles and their write effects in a single typed runtime; modes denote operational roles rather than modal/deontic logic. The contract enforces constraints such as no overwriting evidence, branch isolation, grounded multi-subject timers, and guarded state changes.