English

Aligning Transformers with Weisfeiler-Leman

Machine Learning 2024-06-06 v1

Abstract

Graph neural network architectures aligned with the kk-dimensional Weisfeiler--Leman (kk-WL) hierarchy offer theoretically well-understood expressive power. However, these architectures often fail to deliver state-of-the-art predictive performance on real-world graphs, limiting their practical utility. While recent works aligning graph transformer architectures with the kk-WL hierarchy have shown promising empirical results, employing transformers for higher orders of kk remains challenging due to a prohibitive runtime and memory complexity of self-attention as well as impractical architectural assumptions, such as an infeasible number of attention heads. Here, we advance the alignment of transformers with the kk-WL hierarchy, showing stronger expressivity results for each kk, making them more feasible in practice. In addition, we develop a theoretical framework that allows the study of established positional encodings such as Laplacian PEs and SPE. We evaluate our transformers on the large-scale PCQM4Mv2 dataset, showing competitive predictive performance with the state-of-the-art and demonstrating strong downstream performance when fine-tuning them on small-scale molecular datasets. Our code is available at https://github.com/luis-mueller/wl-transformers.

Keywords

Cite

@article{arxiv.2406.03148,
  title  = {Aligning Transformers with Weisfeiler-Leman},
  author = {Luis Müller and Christopher Morris},
  journal= {arXiv preprint arXiv:2406.03148},
  year   = {2024}
}

Comments

Accepted at ICML 2024

R2 v1 2026-06-28T16:54:20.900Z