Reifying knowledge graphs enables zero-shot link prediction with simple GNNs
Reification as a Transferable Vocabulary: Zero-Shot Link Prediction with Vanilla GNNs
Machine LearningArtificial Intelligence
Summary
Predicting new connections in knowledge graphs is often hard without training on similar data. The authors found that by turning every fact in a graph into its own node connected in a standard way, simple graph neural networks can predict new links even on unseen graphs. This method works without changing the model architecture or needing special training tricks. They also showed it can be applied to databases, helping find relationships without extra information. The code and tests are publicly available for others to use.
What this means in practice
- •For data engineers: Use simple graph neural networks on reified knowledge graphs to predict missing links without retraining on new data.
- •For database administrators: Identify foreign-key relationships in relational databases without depending on schema text or content values via pretrained GNNs.
Authors
Camille Pradel
Abstract
Knowledge graph foundation models such as ULTRA achieve zero-shot link prediction on unseen graphs through dedicated architectures that hard-code a transfer mechanism. In this work we move that mechanism out of the architecture and into the representation, by \emph{reifying} the input graph: every fact becomes a node, connected to its subject, object, and relation type through a fixed vocabulary of six meta-relations, with relation types as anonymous shared nodes rather than model parameters. On this representation, five textbook GNNs (GAT, GINE with sum and with mean+max aggregation, GraphSAGE, R-GCN), each trained on a single knowledge graph of 4,245 triples for 30 minutes on one NVIDIA A100, transfer zero-shot to 40 inductive link-prediction benchmarks. The best of them, an off-the-shelf GAT, matches ULTRA, a dedicated foundation model pretrained on three graphs, across ULTRA's own evaluation suite. The same fixed vocabulary extends to relational databases, a row becoming an entity and a foreign-key column a relation type; a preliminary probe on two unseen databases, with no cell values, schema text or in-context labels, shows a model of this family pretrained on three knowledge graphs ranking foreign-key targets far above random-initialization and degree controls. We release the code, the checkpoints, and the evaluation pipeline for all 40 benchmarks.