Research map / Graph learning / Graph self-supervised learning
Learning from Counterfactual Links 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 |
|---|---|---|---|
| train with counterfactual links and discrepancy regularization in addition to factual links (ablation) | CFLP w/ JKNet · Cora | Hits@20: 58.58$\pm$0.23 -> 65.57$\pm$1.05 | Helped |
| add counterfactual link loss (ablation) | CFLP w/ JKNet · Cora | AUC: 89.16$\pm$0.93 -> 93.05$\pm$0.24 | Helped |
| add discrepancy loss (ablation) | CFLP w/ JKNet · Cora | Hits@20: 62.27$\pm$0.84 -> 65.57$\pm$1.05 | Helped |
| add counterfactual link loss (ablation) | CFLP w/ JKNet · CiteSeer | Hits@20: 65.49$\pm$2.18 -> 68.09$\pm$1.49 | Helped |
| add discrepancy loss (ablation) | CFLP w/ JKNet · CiteSeer | AUC: 91.98$\pm$0.17 -> 92.12$\pm$0.47 | Helped |
| use a treatment with weaker association to link existence, such as K-core rather than Katz | CFLP w/ JKNet · Cora | Hits@20: 56.6$\pm$3.4 -> 65.6$\pm$1.1 | Helped |
| use CFLP rather than the strongest baseline | link prediction · OGB-ddi | Hits@20: 60.56$\pm$8.69 -> 86.08$\pm$1.98 | Helped |
| use CFLP rather than the strongest baseline | link prediction · Facebook | AUC: 99.60$\pm$0.02 -> 99.31$\pm$0.04 | Hurt |
| replace MVGRL embeddings with GRACE embeddings for counterfactual matching (ablation) | CFLP w/ JKNet · Cora | Hits@20: 65.57$\pm$1.05 -> 62.54$\pm$1.41 | Hurt |
| replace MVGRL embeddings with DGI embeddings for counterfactual matching (ablation) | CFLP w/ JKNet · CiteSeer | Hits@20: 68.09$\pm$1.49 -> 72.17$\pm$1.08 | Helped |
Closest papers by meaning
- Implicit degree bias in the link prediction task (ICML 2025 · similarity 0.87)
- Accurate Link Prediction for Edge-Incomplete Graphs via PU Learning (AAAI 2025 · similarity 0.86)
- Revisiting Link Prediction: a data perspective (ICLR 2024 · similarity 0.86)
- Bridging Theory and Practice in Link Representation with Graph Neural Networks (NeurIPS 2025 · similarity 0.85)
- Social Recommendation via Graph-Level Counterfactual Augmentation (AAAI 2025 · similarity 0.85)
- Future Link Prediction Without Memory or Aggregation (NeurIPS 2025 · similarity 0.85)
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.