Research map / Label-efficient and robust learning / Domain adaptation and generalization
GP2F: Cross-Domain Graph Prompting with Adaptive Fusion of Pre-trained Graph Neural Networks
Dongxiao He, Wenxuan Sun, Yongqi Huang, Jitao Zhao, Di Jin
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 both contrastive alignment loss and topology-consistent fusion loss (versus merging branches without loss) (ablation) | GP2F · 1-shot node classification (GRACE) | accuracy | Improved |
| add contrastive alignment loss L_ctr (ablation) | GP2F · 1-shot node classification (GRACE) | accuracy | Improved |
| add topology-consistent fusion loss L_fus (ablation) | GP2F · 1-shot node classification (GRACE) | accuracy | Improved |
| add frozen pre-trained branch to the adapted (prompt) branch (ablation) | GP2F · 1-shot node classification (GRACE) | accuracy | Improved |
| larger adapter projection dimension r (ablation) | GP2F · 1-shot node classification (GRACE) | accuracy | Mixed |
| use GP2F instead of best baseline GPL method | GP2F · 1-shot node classification, 7 datasets | accuracy: improves the performance by 1.66% on average | Improved |
| use GP2F instead of best baseline on CiteSeer and CS | GP2F · 1-shot node classification | accuracy: 3.45% and 3.49% | Improved |
| use GP2F instead of best baseline on BZR and COX2 | GP2F · 50-shot graph classification | AUC: 1.83% and 3.02% | Improved |
| use linear probing instead of graph prompt learning | GRACE/DGI/GraphMAE · 1-shot node classification | accuracy | No clear effect |
| pre-train on multiple domains (graph foundation models) instead of single domain | GCOPE/MDGPT/MDGFM · 1-shot cross-domain node classification | accuracy | Hurt |
| use GP2F with DGI or GraphMAE pre-training instead of GRACE | GP2F · 1-shot node classification | accuracy | No clear effect |
Closest papers by meaning
- One Prompt Fits All: Universal Graph Adaptation for Pretrained Models (NeurIPS 2025 · similarity 0.93)
- GraphTOP: Graph Topology-Oriented Prompting for Graph Neural Networks (NeurIPS 2025 · similarity 0.91)
- Spectral Reversal: Counteracting Singular Value Bias for Graph Prompting (NeurIPS 2026 · similarity 0.91)
- HeterGP: Bridging Heterogeneity in Graph Neural Networks with Multi-View Prompting (AAAI 2025 · similarity 0.90)
- ProG: A Graph Prompt Learning Benchmark (NeurIPS 2024 · similarity 0.90)
- Edge Prompt Tuning for Graph Neural Networks (ICLR 2025 · similarity 0.90)
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.
