Research map / Graph learning / GNN architectures
Graph Neural Networks are Dynamic Programmers
Andrew Joseph Dudzik, Petar Veličković
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add order-3 messages and edge updates | MPNN · CLRS six tasks OOD, large models | accuracy: 35.24% -> 39.43% | Improved |
| Add order-3 messages and edge updates | MPNN · CLRS six tasks OOD, small models | accuracy: 33.57% -> 36.23% | Mixed |
| Add order-3 messages and edge updates | PGN · CLRS edge-centric tasks OOD | accuracy: 35.03% -> 39.08%; 4.44% ± 1.06 | Improved |
| Add order-3 messages and edge updates | PGN · CLRS other tasks OOD | accuracy: 35.37% -> 36.33%; 1.01% ± 0.11 | Improved |
| Add order-3 messages and edge updates | PGN · CLRS 27 tasks OOD | accuracy: 35.20% -> 37.70% | Improved |
| Add order-3 messages and edge updates | PGN · Bellman-Ford, CLRS OOD | accuracy: 80.05% ± 0.87 -> 77.69% ± 0.78 | Hurt |
| Add order-3 messages and edge updates | PGN · Floyd-Warshall, CLRS OOD | accuracy: 8.01% ± 1.31 -> 17.31% ± 0.92 | Improved |
| Add order-3 messages and edge updates | PGN · LCS Length, CLRS OOD | accuracy: 52.87% ± 2.35 -> 67.24% ± 4.93 | Improved |
| Add order-3 messages and edge updates | MPNN · Floyd-Warshall, CLRS OOD | accuracy: 7.46% ± 0.63 -> 9.00% ± 0.81 | Improved |
Closest papers by meaning
- Deep Equilibrium Algorithmic Reasoning (NeurIPS 2024 · similarity 0.89)
- Primal-Dual Neural Algorithmic Reasoning (ICML 2025 · similarity 0.88)
- Understanding Transformer Reasoning Capabilities via Graph Algorithms (NeurIPS 2024 · similarity 0.88)
- On the Power of Small-size Graph Neural Networks for Linear Programming (NeurIPS 2024 · similarity 0.87)
- Recurrent Graph Neural Networks and Their Connections to Bisimulation and Logic (AAAI 2024 · similarity 0.87)
- Expressive Power of Graph Neural Networks for (Mixed-Integer) Quadratic Programs (ICML 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
