GraphOracle: Efficient Fully-Inductive Knowledge Graph Reasoning via Relation-Dependency Graphs

TL;DR

GraphOracle tackles fully-inductive knowledge graph reasoning, where both entities and relations are unseen at test time, by converting KGs into a sparse, directed Relation-Dependency Graph (RDG) that encodes directional relational dependencies. A query-conditioned multi-head attention over the RDG produces context-aware relation embeddings, which guide a second GNN to perform inductive message passing on the original KG for unseen entities and relations. The model is pre-trained across multiple general-domain KGs and can be fine-tuned with lightweight adaptation on target graphs, achieving state-of-the-art results across 60 benchmarks and showing strong cross-domain generalization; GraphOracle+ further boosts performance by integrating modality-specific external information. This framework demonstrates robust generalization, computational efficiency due to the sparse RDG, and practical potential for scalable cross-domain KG reasoning and transfer learning.

Abstract

Knowledge graph reasoning in the fully-inductive setting, where both entities and relations at test time are unseen during training, remains an open challenge. In this work, we introduce GraphOracle, a novel framework that achieves robust fully-inductive reasoning by transforming each knowledge graph into a Relation-Dependency Graph (RDG). The RDG encodes directed precedence links between relations, capturing essential compositional patterns while drastically reducing graph density. Conditioned on a query relation, a multi-head attention mechanism propagates information over the RDG to produce context-aware relation embeddings. These embeddings then guide a second GNN to perform inductive message passing over the original knowledge graph, enabling prediction on entirely new entities and relations. Comprehensive experiments on 60 benchmarks demonstrate that GraphOracle outperforms prior methods by up to 25% in fully-inductive and 28% in cross-domain scenarios. Our analysis further confirms that the compact RDG structure and attention-based propagation are key to efficient and accurate generalization.

Figures (8)

• Figure 1: Overview of the GraphOracle process that predicts the answer entity $e_a$ from a given query $(e_1, r_1, ?)$: Given a Knowledge Graph, we first construct the Relation Dependency Graph (RDG). Then, a multi-head attention mechanism combined with a GNN is used to propagate messages among RDG to obtain relation representations $\bm h_{r|r_q}^{L_r}$, which are then used in another GNN for message passing over entity representations. Finally, the candidate entities are scored and evaluated based on the aggregated entity representations, and then ranked for answer entity prediction.
• Figure 2: Comparison of the MRR performance (the larger the better) between GraphOracle and supervised SOTA methods across various datasets. Note that Amazon-book uses NDCG@20 due to its adaptation to the recommendation task.
• Figure 3: Perturbation Analysis of RDG Edges by Attention-Derived Importance Scores.
• Figure 4: Impact of Number of Pre-trained Datasets on Zero-Shot Evaluation Metrics.
• Figure 5: Comparison on PrimeKG: Evaluating GraphOracle Enhanced by External Entity information (GraphOracle+).

Theorems & Definitions (8)

• theorem 1: Representation Capacity
• proof
• theorem 2: Expressive Power
• proof
• theorem 3: Generalization Error Bound
• proof
• theorem 4: Inductive Generalization
• proof