SwapGT augments tokenized graph Transformers with a random token swapping operation and a center alignment loss, reporting state-of-the-art accuracy on eight node classification datasets.
How attentive are graph attention networks? In Proceedings of the International Conference on Learning Representations, 2022
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Rethinking Tokenized Graph Transformers for Node Classification
SwapGT augments tokenized graph Transformers with a random token swapping operation and a center alignment loss, reporting state-of-the-art accuracy on eight node classification datasets.