aiqbench analysis
This research proposes a method for training Graph Neural Networks (GNNs) on coarse-grained replicas of massive graphs, enabling zero-shot deployment on full-resolution networks without retraining. By leveraging geometric renormalization to preserve structural similarity across scales, the approach significantly reduces infrastructure costs while maintaining predictive performance in both synthetic and real-world scenarios. The findings suggest that architectural scale-equivariance may be more critical than raw node count for efficient GNN scaling.
Train Small, Deploy Large: Zero-Shot GNN Transfer Through Geometric Renormalization
arXiv:2607.27767v1 Announce Type: new Abstract: Graph neural networks (GNNs) can operate on large graphs but become infrastructure-sensitive at the scale of millions of nodes and typically require scalable training techniques for even larger graphs. This raises a central question: when can a model trained on a smaller, scaled-down replica of a graph be deployed on the full-resolution graph without retraining? We introduce a zero-shot transfer protocol in which a GNN is trained on a graph coarse-grained by geometric renormalization (GR), and the resulting weights are transferred directly to the original network. Across synthetic and real-world networks, training on GR scaled-down replicas preserves much of the original-scale predictive performance while significantly reducing training cost. We further find that learned representations and predictive trajectories remain aligned across scales. These findings suggest that structural similarity may be more important than network size in determining GNN transferability, opening a path toward scale-equivariant graph architectures.
Abstract:Graph neural networks (GNNs) can operate on large graphs but become infrastructure-sensitive at the scale of millions of nodes and typically require scalable training techniques for even larger graphs. This raises a central question: when can a model trained on a smaller, scaled-down replica of a graph be deployed on the full-resolution graph without retraining? We introduce a zero-shot transfer protocol in which a GNN is trained on a graph coarse-grained by geometric renormalization (GR), and the resulting weights are transferred directly to the original network. Across synthetic and real-world networks, training on GR scaled-down replicas preserves much of the original-scale predictive performance while significantly reducing training cost. We further find that learned representations and predictive trajectories remain aligned across scales. These findings suggest that structural similarity may be more important than network size in determining GNN transferability, opening a path toward scale-equivariant graph architectures.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Physics and Society (physics.soc-ph)
arXiv:2607.27767 [cs.LG]
(or arXiv:2607.27767v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2607.27767 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 246 words
Related coverage
Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics
Latent Space · Oct 13, 2026
Runway’s WorldPrompt and the Engineering of Real-Time Worlds
Latent Space · Sep 25, 2026
Graph-Based Inference and Topology-Aware Multi-Agent Reinforcement Learning for Large-Scale Railway Network Management
arXiv · Sep 26, 2026
Learnable composition for neural operators
arXiv · Sep 4, 2026