Research map / Graph learning / GNN architectures
Message Tuning Outshines Graph Prompt Tuning: A Prismatic Space Perspective
Yancheng Chen, Dun Ma, Shuai Zhang, Yang Liu, Xixun Lin, Xiangyu Zhao, Wenguo Yang, Wei Chen, Chuan Zhou
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 MTG message prototypes in every layer instead of input-space graph prompt tuning (best GPF-plus / All-in-one) | graph prompt tuning · 15 datasets, few-shot node and graph classification (GCN, best over pre-training) | accuracy | Improved |
| replace GPF-plus with MTG | GPF-plus · Cora 1-shot | accuracy: 55.77 -> 58.54 | Improved |
| replace fine-tuning with MTG | Fine-tuning · ogbn-arxiv 1-shot | accuracy: 16.21 -> 75.97 | Improved |
| replace All-in-one with MTG | All-in-one · ENZYMES 1-shot | accuracy: 23.96 -> 26.08 | Improved |
| MTG versus GPF-plus at 3-shot on Actor | GPF-plus · Actor 3-shot node classification | accuracy: 43.59 -> 37.62 | Hurt |
| MTG instead of GPF-plus at 5-shot on Texas | GPF-plus · Texas 5-shot node classification | accuracy: 99.12 -> 98.76 | Hurt |
| add MTG to pre-trained GCN across six pre-training strategies | fine-tuning, GPF-plus, All-in-one · 1-shot node and graph classification | negative transfer rate | Improved |
| use MTG instead of fine-tuning under DGI and GraphCL pre-training | Fine-tuning · ogbn-arxiv 1-shot | accuracy | Improved |
| vary number of message prototypes m (ablation) | MTG · Cora 1-shot and BZR 1-shot | accuracy | No clear effect |
| swap GNN backbone among GCN, GAT, GraphSAGE, GIN, GT (ablation) | MTG · Wisconsin 1-shot and PROTEINS 1-shot | accuracy | No clear effect |
Closest papers by meaning
- One Prompt Fits All: Universal Graph Adaptation for Pretrained Models (NeurIPS 2025 · similarity 0.89)
- GraphTOP: Graph Topology-Oriented Prompting for Graph Neural Networks (NeurIPS 2025 · similarity 0.89)
- Edge Prompt Tuning for Graph Neural Networks (ICLR 2025 · similarity 0.89)
- Does Graph Prompt Work? A Data Operation Perspective with Theoretical Analysis (ICML 2025 · similarity 0.88)
- AdapterGNN: Parameter-Efficient Fine-Tuning Improves Generalization in GNNs (AAAI 2024 · similarity 0.88)
- HG-Adapter: Improving Pre-Trained Heterogeneous Graph Neural Networks with Dual Adapters (ICLR 2025 · similarity 0.88)
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.
