Research map / Graph learning / Knowledge and temporal graphs
Neural Bellman-Ford Networks: A General Graph Neural Network Framework for Link Prediction
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use learned message and aggregation operators instead of the handcrafted path operators in DRUM | NBFNet · FB15k-237 and WN18RR | H@1: average relative gain of 21% | Helped |
| Use NBFNet rather than the best embedding baseline | knowledge graph completion · FB15k-237 and WN18RR | H@1: average relative performance gain of 18% | Helped |
| Use NBFNet rather than GraIL | inductive relation prediction · FB15k-237 and WN18RR inductive splits | HITS@10: average relative performance gain of 22% | Helped |
| Use NBFNet instead of S-VGAE | homogeneous graph link prediction · Citeseer | AUROC: 0.947 -> 0.923 | Hurt |
| Replace TransE message with DistMult message, keeping sum aggregation (ablation) | NBFNet · FB15k-237 | MRR: 0.297 -> 0.388 | Helped |
| Replace TransE message with RotatE message, keeping sum aggregation (ablation) | NBFNet · FB15k-237 | MRR: 0.297 -> 0.392 | Helped |
| Replace sum aggregation with PNA aggregation, using DistMult message (ablation) | NBFNet · FB15k-237 | MRR: 0.388 -> 0.415 | Helped |
| Replace sum aggregation with PNA aggregation, using RotatE message (ablation) | NBFNet · FB15k-237 | MRR: +0.022 | Helped |
| Increase depth from 2 to 4 GNN layers (ablation) | NBFNet · FB15k-237 | MRR: 0.345 -> 0.409 | Helped |
| Increase depth from 4 to 6 GNN layers (ablation) | NBFNet · FB15k-237 | MRR: 0.409 -> 0.415 | Helped |
| Increase depth from 6 to 8 GNN layers (ablation) | NBFNet · FB15k-237 | MRR: 0.415 -> 0.416 | No clear effect |
Closest papers by meaning
- GNNs Meet Sequence Models Along the Shortest-Path: an Expressive Method for Link Prediction (NeurIPS 2026 · similarity 0.89)
- Bridging Theory and Practice in Link Representation with Graph Neural Networks (NeurIPS 2025 · similarity 0.88)
- A Topological Perspective on Demystifying GNN-Based Link Prediction Performance (ICLR 2024 · similarity 0.88)
- Improving Graph Neural Networks by Learning Continuous Edge Directions (ICLR 2025 · similarity 0.87)
- NetInfoF Framework: Measuring and Exploiting Network Usable Information (ICLR 2024 · similarity 0.87)
- Beyond Message Passing: Neural Graph Pattern Machine (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.