Recent Advances & Key Papers

Landmark papers, 2020–2026 trends, theses to read, code repos, and open problems.

1. Foundational Papers (Must Read)

YearPaperWhy it matters
2014Bruna - Spectral NetworksFirst spectral GNN - started the field
2016Defferrard - ChebNetLocalized spectral filters
2017Kipf & Welling - GCNThe model everyone starts with
2017Gilmer - MPNNUnified message passing framework
2017Hamilton - GraphSAGEInductive learning at scale
2018Veličković - GATAttention on graphs
2019Xu - GIN / WL powerExpressiveness upper bound

2. Benchmarks & Ecosystem

3. Expressiveness & Limits

4. Graph Transformers Era

5. Self-Supervised & Foundation Models

6. Molecules & Equivariant GNNs

7. LLM + Graph (2024–2026)

Full index: Papers Hub

8. Theses & Books

9. Open Problems

  1. Scaling laws for GNNs - do they follow LLM-style power laws?
  2. True graph foundation models - one pretrained model for any graph domain?
  3. Over-squashing cure - practical fix without $O(n^2)$ attention
  4. Heterophily at scale - production models on web-scale heterophilous graphs
  5. Dynamic graphs + LLMs - streaming KG updates with frozen LLM
  6. Theory of depth - principled layer count selection
  7. Fairness and bias on graph recommendations
  8. Explainability - which subgraph caused the prediction?