Research map / Graph learning / Graph self-supervised learning
LLMs and graph foundation models
Combine language models with graph learning for text-attributed and foundation models.
117 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Graph self-supervised learning, keyword cluster “classification · semi supervised · representation”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 1 of 2 tracked benchmarks.
22 papers in 2023–24 and 74 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 6.4% to 12.7%.
Open questions the data points to
- Citeseer (accuracy): best went from 69.37 (up to 2024) to 72.37 in 2025 — Dynamic Bundling with Large Language Models for Zero-Shot Inference on Text-Attributed Graphs (NeurIPS 2025 · 0 citations)
What looks settled
- WikiCS (accuracy): best up to 2024 78.34, best in 2025–26 71.17 from 4 papers, at most 0.2% better, if at all.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2017 · Deep Generative Models for Relational Data with Side Information (ICML 2017 · 10 citations)
- 2018 · Diffusion Maps for Textual Network Embedding (NeurIPS 2018 · 18 citations)
- 2019 · Graphonomy: Universal Human Parsing via Graph Transfer Learning (CVPR 2019 · 198 citations)
- 2020 · Graph LSTM with Context-Gated Mechanism for Spoken Language Understanding (AAAI 2020 · 54 citations)
- 2021 · OGB-LSC: A Large-Scale Challenge for Machine Learning on Graphs (NeurIPS 2021 · 119 citations)
- 2022 · OrphicX: A Causality-Inspired Latent Variable Model for Interpreting Graph Neural Networks (CVPR 2022 · 54 citations)
- 2023 · GIMLET: A Unified Graph-Text Model for Instruction-Based Molecule Zero-Shot Learning (NeurIPS 2023 · 26 citations)
- 2024 · Adaptive Graph Learning for Multimodal Conversational Emotion Detection (AAAI 2024 · 49 citations)
- 2025 · Leveraging Large Language Models for Node Generation in Few-Shot Learning on Text-Attributed Graphs (AAAI 2025 · 12 citations)
- 2026 · GraphPFN: A Prior-Data Fitted Graph Foundation Model (ICML 2026 · 0 citations)
Earliest papers here
- Deep Generative Models for Relational Data with Side Information (ICML 2017 · 10 citations)
- Diffusion Maps for Textual Network Embedding (NeurIPS 2018 · 18 citations)
- Graphonomy: Universal Human Parsing via Graph Transfer Learning (CVPR 2019 · 198 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
WikiCS · accuracy (higher is better)
6 papers, 2024–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2024 | 78.34 | One For All: Towards Training One Graph Model For All Classification Tasks (ICLR 2024 · 8 citations) |
Citeseer · accuracy (higher is better)
4 papers, 2024–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2024 | 69.37 | Label-free Node Classification on Graphs with Large Language Models (LLMs) (ICLR 2024 · 10 citations) |
| 2025 | 72.37 | Dynamic Bundling with Large Language Models for Zero-Shot Inference on Text-Attributed Graphs (NeurIPS 2025 · 0 citations) |
What has been tried
70 papers here report 777 measured changes (a component added, removed or swapped, compared with a baseline): helped 77%, no clear effect 4%, hurt 15%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add acoustic modality | IEMOCAP · weighted F1 | 65.97 -> 66.88 | Helped | Adaptive Graph Learning for Multimodal Conversational Emotion Detection (AAAI 2024 · 49 citations) |
| Add explanation and prediction features alongside original-text features | ogbn-arxiv test · accuracy | 0.7504 ± 0.0020 -> 0.7750 ± 0.0012 | Helped | Harnessing Explanations: LLM-to-LM Interpreter for Enhanced Text-Attributed Graph Representation Learning (ICLR 2024 · 16 citations) |
| add the edge predictor to connect generated nodes to raw graph nodes | Cora 1-shot · accuracy | 75.12(1.60) -> 76.48(0.90) | Helped | Leveraging Large Language Models for Node Generation in Few-Shot Learning on Text-Attributed Graphs (AAAI 2025 · 12 citations) |
| Apply graph instruction tuning to Llama3 | Cora · accuracy | 14.92 -> 91.51 | Helped | Large Language Model Meets Graph Neural Network in Knowledge Distillation (AAAI 2025 · 11 citations) |
| Replace vanilla zero-shot annotation with one-shot vanilla prompting | Ogbn-products · accuracy | 75.33 ± 4.99 -> 78.67 ± 4.50 | Helped | Label-free Node Classification on Graphs with Large Language Models (LLMs) (ICLR 2024 · 10 citations) |
| Replace the OFA GCN backbone with SGC | Node classification co-training · performance | Helped | Text-space Graph Foundation Models: Comprehensive Benchmarks and New Insights (NeurIPS 2024 · 8 citations) | |
| remove review information | Book and Yelp · recommendation performance | Hurt | Semantic Enhanced Heterogeneous Hypergraph Network for Collaborative Filtering (AAAI 2025 · 4 citations) | |
| add LoRA fine-tuning to the LLM | Cora, PubMed, Arxiv · accuracy | Cora: 0.8319 -> 0.8824; PubMed: 0.8756 -> 0.9213; Arxiv: 0.6513 -> 0.7235 | Helped | Multi-View Empowered Structural Graph Wordification for Language Models (AAAI 2025 · 3 citations) |
| remove graph structure and use multimodal features alone | Goodreads-NC · accuracy | 58.75 ±0.05 | Hurt | Mosaic of Modalities: A Comprehensive Benchmark for Multimodal Graph Learning (CVPR 2025 · 3 citations) |
| Replace the MLP parsing head with GCN refinement | RED140 · Discrete architecture F1 | 75.7 -> 79.5 | Helped | Tracing Hyperparameter Dependencies for Model Parsing via Learnable Graph Pooling Network (NeurIPS 2024 · 3 citations) |
| Add self-retrieval decoding | Twitter · Hits@100 | 63.07 -> 65.08 | Helped | MSR: A Multifaceted Self-Retrieval Framework for Microscopic Cascade Prediction (AAAI 2025 · 2 citations) |
| Replace second-order Myerson-Taylor index with second-order Shapley-Taylor | BA-2Motifs, BA-HouseGrid (GCN) · Explanation accuracy | Hurt | Explaining Graph Neural Networks via Structure-aware Interaction Index (ICML 2024 · 2 citations) |
What reviewers will expect
Datasets and baselines named in the 36 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- Cora · 18 of 36 papers
- PubMed · 16 of 36 papers
- ogbn-arxiv · 8 of 36 papers
- Citeseer · 8 of 36 papers
- WikiCS · 7 of 36 papers
- Arxiv · 5 of 36 papers
- FB15K237 · 5 of 36 papers
- ENZYMES · 3 of 36 papers
Baselines
- GCN · 14 of 36 papers
- GAT · 11 of 36 papers
- LLaGA · 8 of 36 papers
- GraphGPT · 7 of 36 papers
- DGI · 6 of 36 papers
- OFA · 6 of 36 papers
- GraphMAE · 5 of 36 papers
- GraphSAGE · 4 of 36 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 96 papers; from OpenAlex and the papers' affiliations).
- Peking University · 6 papers
- Michigan State University · 4 papers
- Emory University · 4 papers
- Nanyang Technological University · 3 papers
- The Chinese University of Hong Kong · 3 papers
- Beihang University · 3 papers
- Amazon · 3 papers
- University of Wisconsin - Madison · 3 papers
- Carnegie Mellon University · 2 papers
- University of Washington · 2 papers
Use it
Download BibTeX of 38 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Semi-supervised node representation learning (214 papers)
- Heterophily and imbalance on graphs (178 papers)
- Brain and biological graph learning (78 papers)
- Multi-view graph clustering (70 papers)
Elsewhere in the topic
- Diffusion-based graph generation (138 papers)
- Graph anomaly detection (101 papers)
- Molecular graph pretraining (101 papers)
- Graph domain adaptation (82 papers)
- Graph contrastive learning analysis (61 papers)
- Graph out-of-distribution generalization (58 papers)
- Contrastive learning with negatives and views (51 papers)
- Augmentation-free graph learning (25 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).