2025· International Journal of Emerging Trends in Multidisciplinary Research· 0 citations
TL;DR
The proposed framework provides a scalable foundation for graph-based artificial intelligence and has applications in biomedical knowledge discovery, financial fraud detection, industrial digital twins, recommendation systems, cybersecurity intelligence, scientific literature mining, and smart governance.
Abstract
Graph Foundation Models (GFMs) enable universal representation learning from heterogeneous graph-structured data through large-scale self-supervised pretraining. Unlike traditional Graph Neural Networks (GNNs), GFMs learn transferable structural, semantic, and contextual knowledge across multiple domains, including healthcare, finance, manufacturing, cybersecurity, and smart cities, making them highly effective for cross-domain knowledge integration. This paper presents an intelligent multi-layer GFM framework for integrating heterogeneous knowledge graphs and supporting scalable cross-domain analytics. The framework combines graph representation learning, knowledge graph embedding, transformer-based graph encoders, self-supervised contrastive learning, and domain adaptation to perform semantic alignment, feature extraction, graph embedding optimization, and downstream reasoning within a unified environment. A cross-domain integration pipeline automatically aligns entities, relationships, semantics, and graph topologies from diverse data sources while transformer-based graph attention captures both local and global structural dependencies. The proposed framework is evaluated using metrics such as knowledge integration accuracy, graph embedding accuracy, node classification, link prediction, semantic consistency, computational efficiency, scalability, and inference latency. Experimental results demonstrate improved cross-domain representation learning, enhanced transfer learning, reduced feature engineering, and lower dependence on labeled data compared with conventional graph learning approaches. The framework provides a scalable foundation for graph-based artificial intelligence and has applications in biomedical knowledge discovery, financial fraud detection, industrial digital twins, recommendation systems, cybersecurity intelligence, scientific literature mining, and smart governance. Overall, Graph Foundation Models offer a promising solution for universal graph intelligence, enabling accurate cross-domain reasoning, predictive analytics, and explainable decision-making.
Graph Foundation Models (GFMs) have recently emerged as a promising paradigm for enabling knowledge transfer across diverse domains. Unlike traditional graph learning methods that are typically designed for in-domain settings, GFMs aim to learn transferable knowledge that can generalize to unseen graph domains. However, unlike language or visual data, graphs lack intrinsic and unified representation units, such as tokens in language and patches in vision, making it challenging to identify transferable knowledge units for building graph foundation models. Existing graph foundation models mainly focus on mitigating domain discrepancies through feature alignment and structure alignment, while overlooking the exploration of transferable knowledge units underlying graph data. Moreover, these methods generally rely on fixed propagation mechanisms during message passing, overlooking the heterogeneity in propagation patterns, as different edges may exhibit distinct propagation patterns for different feature dimensions. To address these limitations, we propose a Propagation-aware Graph Foundation Model (ProGFM), which regards the propagation relationships between edges and feature dimensions as transferable knowledge units. Through a propagation relationship prototype bank, ProGFM learns cross-domain transferable propagation knowledge, enabling adaptive information aggregation in unseen graph domains. Extensive experiments across various cross-domain transfer scenarios demonstrate that ProGFM possesses strong cross-domain knowledge transfer capability and exhibits superior generalization performance compared with existing methods.
This paper reviews hybrid KG–LLM frameworks for predictive analytics, highlighting graph embeddings, Retrieval-Augmented Generation (RAG), transformer-based reasoning, and contextual embedding fusion to improve prediction accuracy, interpretability, and robustness.
Meena Krishnan· International Journal of Mac...· 0 citations
Graph Foundation Models (GFMs) have recently emerged as a promising paradigm for general-purpose graph learning, aiming to learn reusable knowledge that generalizes across diverse graph domains and downstream tasks, reducing the need for specific model development. Achieving this goal requires reconciling the substantial heterogeneity in node features, graph structures, and semantic information across domains. Among them, heterogeneous node features constitute a fundamental input-level barrier, as their dimensionality and semantics vary substantially across datasets. Existing studies typically project or map heterogeneous node features into a fixed-dimensional space, often implicitly equating dimensional uniformity with effective feature unification. Yet dimensional consistency alone does not ensure that the unified features preserve informative semantics and capture transferable patterns that can support cross-domain knowledge transfer. To bridge this conceptual gap, we distill four desiderata for cross-domain graph feature unification: formal uniformity, cross-domain transferability, information preservation, and backbone compatibility. Guided by these principles, we propose SliGFM, a graph foundation model built upon topology-aware sliding-window feature encoding and generative reconstruction. SliGFM orders feature dimensions by topological smoothness and scans the reordered features with a shared sliding-window feature encoder, transforming heterogeneous features into a common space of ordered fixed-dimensional feature tokens. This formulation enables a smoothness-aware transformer to capture transferable relational patterns among feature tokens within each node, while the generative reconstruction objective encourages preservation of the original feature information.
Dongxiao He, Siqi Liu, Jitao Zhao et al.· 0 citations
Graph-based data engineering has become a powerful approach for managing and analyzing highly interconnected data across enterprise systems, IoT, social media, healthcare, finance, and scientific domains. Unlike traditional relational databases, graph-based models represent data as interconnected nodes and edges, enabling efficient relationship analysis, semantic understanding, and knowledge discovery. This paper surveys recent advances in graph databases, knowledge graphs, graph neural networks (GNNs), and distributed graph analytics, and proposes an integrated framework for scalable graph construction, semantic enrichment, graph analytics, and AI-driven knowledge extraction. The framework emphasizes scalability, semantic consistency, explainable AI, and continuous graph evolution. Experimental evaluation demonstrates improved relationship discovery, query performance, and knowledge extraction compared with conventional relational approaches, making the proposed framework suitable for intelligent applications in healthcare, cybersecurity, finance, smart manufacturing, and enterprise knowledge management.
Mahabala H.N· International Journal of Dat...· 0 citations
To address the problems of heterogeneous structures, weak semantic alignment, sparse relations, and reasoning conflicts in multi-modal knowledge graph construction, this paper proposes a multi-modal data-driven knowledge graph reasoning model integrating unified encoding, graph representation learning, and rule constraints. Text descriptions, image labels, temporal attributes, and spatial coordinates are mapped into a shared representation space through modality-specific encoders and feature projection matrices. A multi-relational knowledge graph is then constructed through entity extraction, relation identification, triplet generation, and graph storage. On this basis, graph representation learning is introduced to aggregate neighborhood information and mine implicit relations among sparse nodes. To reduce semantic drift and invalid link prediction, temporal order, spatial inclusion, entity dependency, and event consistency rules are embedded into the reasoning score function. Candidate filtering and conflict resolution are further designed to improve inference efficiency under large-scale triplet conditions. Experimental results show that the proposed model achieves 0.894 Precision, 0.872 Recall, 0.883 F1, 0.907 MRR, and 0.931 Hits@10, outperforming TransE, GCN, GAT, R-GCN, and CompGCN. When the candidate triplet scale reaches 10,000, the inference response time of the proposed model is 146 ms, indicating better relation completion accuracy, reasoning stability, and inference efficiency in multi-modal knowledge graph reasoning tasks.
Muhan Zhang· International Conference on...· 0 citations
Graph Domain Incremental Learning (GDIL) aims to acquire knowledge from a continuous stream of graph domains while mitigating catastrophic forgetting. While parameter-isolation methods leveraging graph parameter-efficient adaptation show promise, prompt-based techniques struggle to adapt to GDIL, and low-rank adaptation methods based on a shared classification layer lead to knowledge confusion.Our empirical observations reveal that transferable knowledge is primarily concentrated in the representation layer. Further, we argue that domain-agnostic representations that are not tied to the classification characteristics are needed to assist the new model in capturing more discriminative features for graph domain incremental learning.Motivated by these insights, we propose COllabOrative Knowledge Extraction and integRation (COOKER) method for GDIL to mine inter-domain relationships and uncover the potential of domain-agnostic representations. Specifically, COOKER employs domain-specific LoRA modules and classifiers to capture specific knowledge. A domain-agnostic LoRA module is instantiated to extract transferable knowledge through contrastive acquisition and topology alignment. We introduce collaborative dynamic integration of dual representations to enable adaptive integration, guided by a complementarity loss to eliminate information redundancy. Extensive experiments demonstrate that COOKER significantly outperforms existing baselines, achieving up to a 4.7% improvement in average performance.
Jialu Li, Yu Wang, Wanyu Lin et al.· Proceedings of the 32nd ACM...· 0 citations
