Global Communication or Graph-Specific Memory?
Abstract
Scalable Graph Transformers are commonly trained and evaluated on static large graphs in a transductive setup. Many scalable Graph Transformer components can be formulated as a constant-size shared memory, similar to virtual nodes, providing compressed information about the whole graph. The counterpart of these models in language models and other domains is justified as the input changes, and this mechanism learns to compress some useful information about the input. In transductive learning on a single fixed graph, however, any shared memory can be seen as a constant at test time. This raises the question of what exactly this shared memory does in this static setup. We give preliminary evidence that optimizing a shared memory directly performs similarly to global communication methods, and so normal local message-passing models can embed similar information in their weights. Thus, these settings may be a poor fit for evaluating global communication in graph neural networks.
Related
- Graph Neural Networks for Resource Allocation in Integrated Sensing and Communication
- TRACE: A spatiotemporal contact memory graph network simulator for granular dynamics
- Local-to-Global Theorems Based on Cones with Applications to Graph Coloring
- Stability of strong global and exponential attractors for semilinear beam equations with fractional damping and memory
- Integrating multi-source local information for global undirected graph reconstruction via integer programming
BibTeX
@misc{shirzad2026global,
title = {{Global Communication or Graph-Specific Memory?}},
author = {Hamed Shirzad and Danica J. Sutherland},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.05874},
eprint = {2610.05874},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.05874},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Global Communication or Graph-Specific Memory?
AU - Hamed Shirzad
AU - Danica J. Sutherland
PY - 2026
DA - 2026/10/05
PB - arXiv
DO - 10.48550/arxiv.2610.05874
UR - https://arxiv.org/abs/2610.05874
AB - Scalable Graph Transformers are commonly trained and evaluated on static large graphs in a transductive setup. Many scalable Graph Transformer components can be formulated as a constant-size shared memory, similar to virtual nodes, providing compressed information about the whole graph. The counterpart of these models in language models and other domains is justified as the input changes, and this mechanism learns to compress some useful information about the input. In transductive learning on a single fixed graph, however, any shared memory can be seen as a constant at test time. This raises the question of what exactly this shared memory does in this static setup. We give preliminary evidence that optimizing a shared memory directly performs similarly to global communication methods, and so normal local message-passing models can embed similar information in their weights. Thus, these settings may be a poor fit for evaluating global communication in graph neural networks.
ER -