GraphARC is a scalable benchmark for few-shot graph transformation learning that exposes a comprehension-execution gap in language models on abstract reasoning tasks.
Graph founda- tion models: A comprehensive survey.arXiv preprint arXiv:2505.15116
14 Pith papers cite this work. Polarity classification is still indexing.
14
Pith papers citing it
abstract
Graph-structured data pervades domains such as social networks, biological systems, knowledge graphs, and recommender systems. While foundation models have transformed natural language processing, vision, and multimodal learning through large-scale pretraining and generalization, extending these capabilities to graphs -- characterized by non-Euclidean structures and complex relational semantics -- poses unique challenges and opens new opportunities. To this end, Graph Foundation Models (GFMs) aim to bring scalable, general-purpose intelligence to structured data, enabling broad transfer across graph-centric tasks and domains. This survey provides a comprehensive overview of GFMs, unifying diverse efforts under a modular framework comprising three key components: backbone architectures, pretraining strategies, and adaptation mechanisms. We categorize GFMs by their generalization scope -- universal, task-specific, and domain-specific -- and review representative methods, key innovations, and theoretical insights within each category. Beyond methodology, we examine theoretical foundations including transferability and emergent capabilities, and highlight key challenges such as structural alignment, heterogeneity, scalability, and evaluation. Positioned at the intersection of graph learning and general-purpose AI, GFMs are poised to become foundational infrastructure for open-ended reasoning over structured data. This survey consolidates current progress and outlines future directions to guide research in this rapidly evolving field. Resources are available at https://github.com/Zehong-Wang/Awesome-Foundation-Models-on-Graphs.
citation-role summary
background 3
citation-polarity summary
roles
background 3polarities
background 3representative citing papers
GRL-Safety benchmark shows that safety in graph representation learning depends on interactions between method design and specific graph stresses rather than broad method families.
Graphlets mined as structural tokens improve zero-shot inductive and transductive link prediction in knowledge graph foundation models across 51 diverse graphs.
Frozen embeddings from a pretrained heterogeneous graph transformer over a 6.9M-node metabolic-engineering knowledge graph predict fermentation titers at R²=0.41, outperforming tabular baselines (R²=0.24).
OpenRFM combines a relational transformer backbone with a batch-level ICL layer and homophily-aware synthetic-plus-real pre-training to improve relational in-context learning by ~30% over prior open models and surpass KumoRFMv1.
Traditional ML and deep learning are tied in accuracy on 72 dynamic PSN datasets for protein structure classification, with DL over 10 times slower.
UniGraphLM uses a multi-domain multi-task GNN encoder and adaptive alignment to create unified graph tokens for LLMs across diverse domains and tasks.
Introduces curvature-stratified evaluation showing relational learning model rankings are stable within curvature regimes but shift across them, making performance geometry-dependent.
SPG is a graph foundation model using spectral decomposition via Chebyshev filters and Gromov-Wasserstein prototypes for improved cross-graph transferability.
DNSD replaces the sheaf Laplacian with a sheaf adjacency operator, adds normalization and gating, and empirically outperforms GNN and NSD baselines by up to 30 percentage points on synthetic long-range graph tasks while also improving on real-world benchmarks.
SCGFM creates transferable graph representations by aligning heterogeneous topologies to shared learnable geometric bases via Gromov-Wasserstein distances and re-encoding features accordingly.
G-Defense builds claim-centered graphs from sub-claims, applies RAG for evidence and competing explanations, then uses graph inference to detect fake news veracity and generate intuitive explanation graphs, claiming SOTA results.
QUIET is a hierarchical RVQ-based graph tokenizer with a learned level-weighting gate; it improves several benchmarks but not consistently against the strongest baselines.
The paper synthesizes three synergies between LLMs and graphs—augmented retrieval/reasoning, bidirectional KG integration, and graph-enhanced agents—plus LLM uses in graph data management and ML.
citing papers explorer
-
GraphARC: A Comprehensive Benchmark for Graph-Based Abstract Reasoning
GraphARC is a scalable benchmark for few-shot graph transformation learning that exposes a comprehension-execution gap in language models on abstract reasoning tasks.
-
On the Safety of Graph Representation Learning
GRL-Safety benchmark shows that safety in graph representation learning depends on interactions between method design and specific graph stresses rather than broad method families.
-
Graphlets as Building Blocks for Structural Vocabulary in Knowledge Graph Foundation Models
Graphlets mined as structural tokens improve zero-shot inductive and transductive link prediction in knowledge graph foundation models across 51 diverse graphs.
-
Canopy: A Heterograph Foundation Model for Metabolic Engineering
Frozen embeddings from a pretrained heterogeneous graph transformer over a 6.9M-node metabolic-engineering knowledge graph predict fermentation titers at R²=0.41, outperforming tabular baselines (R²=0.24).
-
OpenRFM: Dissecting Relational In-Context Learning
OpenRFM combines a relational transformer backbone with a batch-level ICL layer and homophily-aware synthetic-plus-real pre-training to improve relational in-context learning by ~30% over prior open models and surpass KumoRFMv1.
-
Traditional machine learning vs. deep learning from dynamic graph representations of proteins' 3D folds in the task of protein structure classification
Traditional ML and deep learning are tied in accuracy on 72 dynamic PSN datasets for protein structure classification, with DL over 10 times slower.
-
A Unified Graph Language Model for Multi-Domain Multi-Task Graph Alignment Instruction Tuning
UniGraphLM uses a multi-domain multi-task GNN encoder and adaptive alignment to create unified graph tokens for LLMs across diverse domains and tasks.
-
The Post-GCN Decade Revisited: Curvature-Stratified Evaluation of Relational Learning
Introduces curvature-stratified evaluation showing relational learning model rankings are stable within curvature regimes but shift across them, making performance geometry-dependent.
-
A Graph Foundation Model with Spectral Parsing and Prototype-Guided Spatial Propagation
SPG is a graph foundation model using spectral decomposition via Chebyshev filters and Gromov-Wasserstein prototypes for improved cross-graph transferability.
-
Deep Neural Sheaf Diffusion
DNSD replaces the sheaf Laplacian with a sheaf adjacency operator, adds normalization and gating, and empirically outperforms GNN and NSD baselines by up to 30 percentage points on synthetic long-range graph tasks while also improving on real-world benchmarks.
-
Structure-Centric Graph Foundation Model via Geometric Bases
SCGFM creates transferable graph representations by aligning heterogeneous topologies to shared learnable geometric bases via Gromov-Wasserstein distances and re-encoding features accordingly.
-
A Graph-Enhanced Defense Framework for Explainable Fake News Detection with LLM
G-Defense builds claim-centered graphs from sub-claims, applies RAG for evidence and competing explanations, then uses graph inference to detect fake news veracity and generate intuitive explanation graphs, claiming SOTA results.
-
A Hierarchical Quantized Tokenization Framework for Task-Adaptive Graph Representation Learning
QUIET is a hierarchical RVQ-based graph tokenizer with a learned level-weighting gate; it improves several benchmarks but not consistently against the strongest baselines.
-
LLMs+Graphs: Toward Graph-Native, Synergistic AI Systems
The paper synthesizes three synergies between LLMs and graphs—augmented retrieval/reasoning, bidirectional KG integration, and graph-enhanced agents—plus LLM uses in graph data management and ML.