Graph self-supervised learning: research map
1,274 accepted papers on Graph self-supervised learning in Graph learning, from ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026), grouped into 4 clusters and 13 approaches. The busiest year so far is 2026.
Within Graph learning, its share grew from 27.0% in 2023–24 to 37.1% in 2025–26 (344 → 584 papers at ICML, NeurIPS, CVPR and AAAI, the venues with data for all four years).
Explore Graph self-supervised learning in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
Approaches and key papers
classification · semi supervised · representation · 657 papers
Approaches in this cluster:
- Semi-supervised node representation learning (214 papers)
Learn node embeddings and classifiers on graphs with GCNs and inductive methods. - Heterophily and imbalance on graphs (178 papers)
Handle heterophilic, imbalanced, and heterogeneous graphs in node learning. - LLMs and graph foundation models (117 papers)
Combine language models with graph learning for text-attributed and foundation models. - Brain and biological graph learning (78 papers)
Apply graph learning to brain networks and biological data. - Multi-view graph clustering (70 papers)
Cluster using multi-view graphs with anchor and contrastive representations.
Most cited and most cited since 2024:
- Semi-Supervised Classification with Graph Convolutional Networks (ICLR 2017 · 7,975 citations)
- Inductive Representation Learning on Large Graphs (NeurIPS 2017 · 4,474 citations)
- Adaptive Graph Learning for Multimodal Conversational Emotion Detection (AAAI 2024 · 49 citations)
- Gramformer: Learning Crowd Counting via Graph-Modulated Transformer (AAAI 2024 · 45 citations)
detection · anomaly · ood · 241 papers
Approaches in this cluster:
- Graph domain adaptation (82 papers)
Transfer knowledge across graphs with spectral regularization and foundation models. - Graph anomaly detection (101 papers)
Detect anomalous nodes and graphs with generalist, generative, and diffusion-based methods. - Graph out-of-distribution generalization (58 papers)
Learn invariant graph representations for out-of-distribution generalization and detection.
Most cited and most cited since 2024:
- Open Graph Benchmark: Datasets for Machine Learning on Graphs (NeurIPS 2020 · 484 citations)
- SIGMA: Semantic-Complete Graph Matching for Domain Adaptive Object Detection (CVPR 2022 · 232 citations)
- Revisiting Graph-Based Fraud Detection in Sight of Heterophily and Spectrum (AAAI 2024 · 57 citations)
- ADA-GAD: Anomaly-Denoised Autoencoders for Graph Anomaly Detection (AAAI 2024 · 52 citations)
molecular · generation · molecules · 239 papers
Approaches in this cluster:
- Diffusion-based graph generation (138 papers)
Generate graphs with discrete and score-based diffusion and flow matching. - Molecular graph pretraining (101 papers)
Pretrain molecular graphs with motifs and fragments for property prediction and design.
Most cited and most cited since 2024:
- Junction Tree Variational Autoencoder for Molecular Graph Generation (ICML 2018 · 830 citations)
- Protein Interface Prediction using Graph Convolutional Networks (NeurIPS 2017 · 557 citations)
- Prot2Text: Multimodal Protein’s Function Generation with GNNs and Transformers (AAAI 2024 · 41 citations)
- BindGPT: A Scalable Framework for 3D Molecular Design via Language Modeling and Reinforcement Learning (AAAI 2025 · 11 citations)
graph contrastive · gcl · augmentation · 137 papers
Approaches in this cluster:
- Graph contrastive learning analysis (61 papers)
Study and scale graph contrastive learning with theory and efficient objectives. - Contrastive learning with negatives and views (51 papers)
Improve graph contrastive learning via hard negatives, self-augmentation, and clustering. - Augmentation-free graph learning (25 papers)
Replace or learn graph augmentations, including spectral approaches.
Most cited and most cited since 2024:
- Graph Contrastive Learning with Augmentations (NeurIPS 2020 · 859 citations)
- Augmentation-Free Self-Supervised Learning on Graphs (AAAI 2022 · 189 citations)
- GAMC: An Unsupervised Method for Fake News Detection Using Graph Autoencoder with Masking (AAAI 2024 · 55 citations)
- Propagation Tree Is Not Deep: Adaptive Graph Contrastive Learning Approach for Rumor Detection (AAAI 2024 · 42 citations)
Related topics in Graph learning
- Knowledge and temporal graphs (974)
- Graph algorithms and clustering (817)
- GNN architectures (1,471)