RECORD
Towards Graph Foundation Models: A Transferability Perspective
Machine Learning · Submitted 12 Mar 2025
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim To bridge the gap, we present the first comprehensive taxonomy that categorizes and analyzes existing GFMs through the lens of transferability, structuring GFMs around their application scope (domain-specific vs. general-purpose) and their approaches to knowledge acquisition and transfer.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
In recent years, Graph Foundation Models (GFMs) have gained significant attention for their potential to generalize across diverse graph domains and tasks. Some works focus on Domain-Specific GFMs, which are designed to address a variety of tasks within a specific domain, while others aim to create General-Purpose GFMs that extend the capabilities of domain-specific models to multiple domains. Regardless of the type, transferability is crucial for applying GFMs across different domains and tasks. However, achieving strong transferability is a major challenge due to the structural, feature, and distributional variations in graph data. To date, there has been no systematic research examining and analyzing GFMs from the perspective of transferability. To bridge the gap, we present the first comprehensive taxonomy that categorizes and analyzes existing GFMs through the lens of transferability, structuring GFMs around their application scope (domain-specific vs. general-purpose) and their approaches to knowledge acquisition and transfer. We provide a structured perspective on current progress and identify potential pathways for advancing GFM generalization across diverse graph datasets and tasks. We aims to shed light on the current landscape of GFMs and inspire future research directions in GFM development.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | In recent years, Graph Foundation Models (GFMs) have gained significant attention for their potential to generalize across diverse graph domains and tasks. | background 1.00 | 0.04 | 0.02 | asserted 0.94 |
| s1 | Some works focus on Domain-Specific GFMs, which are designed to address a variety of tasks within a specific domain, while others aim to create General-Purpose GFMs that extend the capabilities of domain-specific models to multiple domains. | background 1.00 | 0.04 | 0.03 | asserted 1.00 |
| s2 | Regardless of the type, transferability is crucial for applying GFMs across different domains and tasks. | background 1.00 | 0.05 | 0.05 | asserted 1.00 |
| s3 | However, achieving strong transferability is a major challenge due to the structural, feature, and distributional variations in graph data. | background 0.94 | 0.05 | 0.04 | asserted 1.00 |
| s4 | To date, there has been no systematic research examining and analyzing GFMs from the perspective of transferability. | gap 1.00 | 0.13 | 0.06 | asserted 1.00 |
| s5 | To bridge the gap, we present the first comprehensive taxonomy that categorizes and analyzes existing GFMs through the lens of transferability, structuring GFMs around their application scope (domain-specific vs. general-purpose) and their approaches to knowledge acquisition and transfer. | contribution 1.00 | 0.95 | 0.27 | asserted 1.00 |
| s6 | We provide a structured perspective on current progress and identify potential pathways for advancing GFM generalization across diverse graph datasets and tasks. | contribution 0.94 | 0.87 | 0.22 | asserted 0.99 |
| s7 | We aims to shed light on the current landscape of GFMs and inspire future research directions in GFM development. | future work 0.86 | 0.26 | 0.11 | asserted 0.56 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
In recent years, Graph Foundation Models (GFMs) have gained significant attention for their potential to generalize across diverse graph domains and tasks. Some works focus on Domain-Specific GFMs, which are designed to address a variety of tasks within a specific domain, while others aim to create General-Purpose GFMs that extend the capabilities of domain-specific models to multiple domains. Regardless of the type, transferability is crucial for applying GFMs across different domains and tasks. However, achieving strong transferability is a major challenge due to the structural, feature, and distributional variations in graph data. To date, there has been no systematic research examining and analyzing GFMs from the perspective of transferability. To bridge the gap, we present the first comprehensive taxonomy that categorizes and analyzes existing GFMs through the lens of transferability, structuring GFMs around their application scope (domain-specific vs. general-purpose) and their approaches to knowledge acquisition and transfer. We provide a structured perspective on current progress and identify potential pathways for advancing GFM generalization across diverse graph datasets and tasks. We aims to shed light on the current landscape of GFMs and inspire future research directions in GFM development.
Human review
References
How the literature uses it
1 arXiv paper cite this in 2 sentences
Papers citing it
Cites it as background 1 paper
-
Recently, there has been a lot of interest in developing graph foundation models (GFMs) — models that after large-scale pretraining can be applied to diverse graph datasets without or with minimal fine-tuning [ref] .
GraphLand: Evaluating Graph Machine Learning Models on Diverse Industrial Data · Introduction · 1.00
What it cites · 84 sentences without a confident label
Uses the method or tool of 1 paper
-
It uses SentenceBERT [ref] to encode both node attributes and class descriptions into a shared semantic space.
Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks · Using PLM for semantic alignment. · 0.93
Cites as support 9 papers
-
For example, BooG [ref] , AnyGraph [ref] , Unigraph [ref] , OFA [ref] , and ProG [ref] unify the structural and feature characteristics from diverse graph data, enabling them to handle tasks across various domains, from social network analysis to chemical molecule prediction.
AnyGraph: Graph Foundation Model in the Wild · Unified Taxonomy of GFMs · 0.97
-
For example, BooG [ref] , AnyGraph [ref] , Unigraph [ref] , OFA [ref] , and ProG [ref] unify the structural and feature characteristics from diverse graph data, enabling them to handle tasks across various domains, from social network analysis to chemical molecule prediction.
One for All: Towards Training One Graph Model for All Classification Tasks · Unified Taxonomy of GFMs · 0.97
-
For example, BooG [ref] , AnyGraph [ref] , Unigraph [ref] , OFA [ref] , and ProG [ref] unify the structural and feature characteristics from diverse graph data, enabling them to handle tasks across various domains, from social network analysis to chemical molecule prediction.
UniGraph: Learning a Unified Cross-Domain Foundation Model for Text-Attributed Graphs · Unified Taxonomy of GFMs · 0.97
Cites as background 15 papers
-
LLM-based models [ref] use large language models to process graph data by converting graphs into natural language-like sequences.
GPT4Graph: Can Large Language Models Understand Graph Structured Data ? An Empirical Evaluation and Benchmarking · LLM-based Models. · 0.90
-
LLM-based models [ref] use large language models to process graph data by converting graphs into natural language-like sequences.
2310.01436 · LLM-based Models. · 0.90
-
LLM-based models [ref] use large language models to process graph data by converting graphs into natural language-like sequences.
Disentangled Representation Learning with Large Language Models for Text-Attributed Graphs · LLM-based Models. · 0.90
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.