Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Lost in Tokenization: Fundamental Trade-offs in Graph Tokenization for Transformers

About

Transformers have become a central architecture for graph learning, but their application to graphs requires first choosing a tokenization: a graph-to-token map that determines which structural information is exposed at the input. In this work, we show that this choice is a fundamental component of transformer expressivity. We examine three tokenizations that serve as building blocks for many existing graph tokenizations: spectral, random-walk, and adjacency tokenizations. We prove that different tokenizations induce distinct depth regimes: the same graph computation may be realizable by a shallow transformer under one tokenization, while requiring substantially larger depth under another. For example, we prove that random-walk tokenization is lossy for any walk length, making it impossible in general to recover the graph from it, and that while spectral tokenization is lossless, it is ill-conditioned for local tasks. We further show that although both random-walk and spectral tokenizations are derived from adjacency information, it is impossible for a limited-depth transformer to convert between tokenization families in general. In particular, we establish lower bounds and impossibility results showing that unfavorable tokenizations may preclude the efficient recovery of more suitable structural representations. Finally, we complement our theory with controlled experiments on synthetic and real-world tasks, validating the predicted separations and showing that different tasks favor different structural views, and combining complementary tokenizations allows the transformer to leverage distinct signals from each representation.

Maya Bechler-Speicher, Gilad Yehudai, Gil Harari, Clayton Sanford, Amir Globerson, Joan Bruna• 2026

Related benchmarks

TaskDatasetResultRank
Graph ClassificationHIV
ROC-AUC0.7644
155
Graph RegressionZINC
MAE0.342
144
Graph property predictionBACE
ROC AUC75.94
111
Graph ClassificationBBBP
ROC-AUC70.14
18
Graph ClassificationTox21
ROC-AUC81.36
10
Node ClassificationMaxClq
F1 Score26.73
10
Node RegressionTopoOrd
MAE0.192
10
Graph RegressionEC-5
RSE0.051
10
Showing 8 of 8 rows

Other info

Follow for update