Research map / Graph learning / GNN architectures
GOAT: A Global Transformer on Large-scale Graphs
Kezhi Kong, Jiuhai Chen, John Kirchenbauer, Renkun Ni, C. Bayan Bruss, Tom Goldstein
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 |
|---|---|---|---|
| adding the local attention module (GOAT vs local-only removed global) (ablation) | GOAT · arxiv-year | validation accuracy: ~3% boost | Improved |
| removing the global attention module (local-only) (ablation) | GOAT · ogbn-arxiv | validation accuracy | No clear effect |
| increasing codebook size from 512 to 4k (ablation) | GOAT · arxiv-year | accuracy: 50.90 to 53.57 | Improved |
| increasing codebook size from 512 to 4k (ablation) | GOAT · ogbn-arxiv | accuracy: 71.92 to 72.41 | Improved |
| layer norm instead of batch norm (ablation) | GOAT · ogbn-arxiv | validation accuracy | No clear effect |
| removing normalization entirely (ablation) | GOAT · ogbn-arxiv and arxiv-year | validation accuracy | Hurt |
| GOAT global-only variant vs GAT-NS on training speed | GOAT-Global-only · ogbn-products | seconds per epoch (batch 1k): 5.58 vs 127.42 for GAT | Improved |
| full GOAT (local + global) vs GAT-NS on training time | GOAT · ogbn-products | seconds per epoch (batch 1k): 109.51 vs 127.42 for GAT | Mixed |
| GOAT vs LINKX/MixHop baselines on heterophilous and homophilous data | GOAT · snap-patents (random 50%) | accuracy: 54.97 vs LINKX 60.12 | Mixed |
Later papers that compared with this method
5 later accepted papers report GOAT as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- GT-SNT: A Linear-Time Transformer for Large-Scale Graphs via Spiking Node Tokenization (AAAI 2026 · better on 2 of 3 shared results)
- UniGoal: Towards Universal Zero-shot Goal-oriented Navigation (CVPR 2025 · better on 2 of 2 shared results)
- Beyond Message Passing: Neural Graph Pattern Machine (ICML 2025 · better on 1 of 1 shared result)
- HeroFilter: Adaptive Spectral Graph Filter for Varying Heterophilic Relations (NeurIPS 2025 · better on 1 of 1 shared result)
- Polynormer: Polynomial-Expressive Graph Transformer in Linear Time (ICLR 2024 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Can Classic GNNs Be Strong Baselines for Graph-level Tasks? Simple Architectures Meet Excellence (ICML 2025 · similarity 0.89)
- Towards Principled Graph Transformers (NeurIPS 2024 · similarity 0.89)
- Classic GNNs are Strong Baselines: Reassessing GNNs for Node Classification (NeurIPS 2024 · similarity 0.89)
- Less is More: on the Over-Globalizing Problem in Graph Transformers (ICML 2024 · similarity 0.89)
- A Scalable and Effective Alternative to Graph Transformers (AAAI 2025 · similarity 0.89)
- What Improves the Generalization of Graph Transformers? A Theoretical Dive into the Self-attention and Positional Encoding (ICML 2024 · similarity 0.89)
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.
