Research map / Graph learning / GNN architectures
Reliable and Compact Graph Fine-Tuning via Graph Sparse Prompting
Bo Jiang, Hao Wu, Beibei Wang, Jin Tang, Bin Luo
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 l1 sparsity on the prompt vector (GSFP) | GPF · Cora 1-shot node classification (GraphMAE) | accuracy: 35.15 -> 38.56 | Improved |
| add l2,1 row sparsity on the prompt matrix (GSmFP) | GPF-plus · Photo 1-shot node classification (GraphMAE) | accuracy: 46.41 -> 49.99 | Improved |
| add l1 sparsity on the prompt vector (GSFP) | GPF · Ogbn-arxiv 1-shot node classification (GraphMAE) | accuracy: 9.11 -> 10.81 | Improved |
| add l1 sparsity on the prompt vector (GSFP) | GPF · PubMed 1-shot node classification (GraphCL) | accuracy: 48.25 -> 53.56 | Improved |
| add l1 sparsity on the prompt vector (GSFP) | GPF · Cora 1-shot node classification (GraphCL) | accuracy: 33.24 -> 33.20 | No clear effect |
| add l2,1 row sparsity on the prompt matrix (GSmFP) | GPF-plus · CiteSeer 1-shot node classification (GraphMAE) | accuracy: 31.05 -> 33.61 | Improved |
| add l2,1 row sparsity on the prompt matrix (GSmFP) | GPF-plus · IMDB-B 1-shot graph classification (GraphCL) | accuracy: 61.74 -> 65.05 | Improved |
| add l1 sparsity on the prompt vector (GSFP) | GPF · IMDB-M 1-shot graph classification (GraphCL) | accuracy: 37.77 -> 38.19 | No clear effect |
| add l1 sparsity on the prompt vector (GSFP) | GPF · Cora with GAT backbone (GraphMAE) | accuracy: 35.18 -> 38.79 | Improved |
| add l2,1 row sparsity on the prompt matrix (GSmFP) | GPF-plus · Photo with GraphSAGE backbone (GraphMAE) | accuracy: 45.55 -> 48.08 | Improved |
| increase sparsity penalty lambda beyond a moderate range (ablation) | GSFP and GSmFP · node classification datasets | accuracy | Hurt |
| use moderate sparsity penalty lambda over lambda=0 (GPF) (ablation) | GSFP and GSmFP · node classification datasets | accuracy | Improved |
Closest papers by meaning
- GraphTOP: Graph Topology-Oriented Prompting for Graph Neural Networks (NeurIPS 2025 · similarity 0.90)
- Edge Prompt Tuning for Graph Neural Networks (ICLR 2025 · similarity 0.89)
- ProG: A Graph Prompt Learning Benchmark (NeurIPS 2024 · similarity 0.88)
- Two Heads Are Better Than One: Boosting Graph Sparse Training via Semantic and Topological Awareness (ICML 2024 · similarity 0.88)
- HeterGP: Bridging Heterogeneity in Graph Neural Networks with Multi-View Prompting (AAAI 2025 · similarity 0.87)
- One Prompt Fits All: Universal Graph Adaptation for Pretrained Models (NeurIPS 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.
