Generative Representation Learning on Hyper-relational Knowledge Graphs via Masked Discrete Diffusion

Hyper-relational knowledge graphs (HKGs) effectively represent complex facts. While inferring new knowledge in HKGs is a critical problem, current methods cast it as a simple link prediction, assuming that nearly all entities and relations within a fact are known, leaving only a single blank to be filled. However, this restricted assumption may not hold in real-world scenarios in which multiple, or even all, constituent components of a fact may be missing simultaneously. To bridge this gap, we introduce a task called fact generation: generating a valid hyper-relational fact from an arbitrarily masked query, i.e., completing a partially observed fact or generating a fact from scratch. We propose KREPE, the first generative representation learning method for HKGs that learns to model the probability distributions of missing components conditioned on the local fact components and global structure of HKGs via a masked discrete diffusion. KREPE models both the intra-fact dependencies by contextual message passing and inter-fact correlations by aggregating stochastically sampled contexts. KREPE seamlessly unifies link prediction and fact generation within a single training framework, achieving state-of-the-art performance on standard HKG link prediction benchmarks and outperforming LLM-based baselines in generating novel and correct facts.

Hyper-relational knowledge graphs (HKGs) represent human knowledge as triplets with auxiliary qualifiers. To infer new knowledge in HKGs, existing methods typically assume that nearly all entities and relations within an unknown fact are known, leaving only a single blank to be filled. However, this assumption is often unrealistic in real-world scenarios, where multiple components of a fact may be missing simultaneously, or even the entire fact may be unknown. We introduce fact generation, the task of generating valid hyper-relational facts from arbitrarily masked queries, including fully masked ones. To address this task, we propose KREPE, the first generative representation learning method for HKGs based on masked discrete diffusion. KREPE models the joint probability distribution of missing entities and relations by capturing both the local dependencies within a fact and the global structural patterns of the HKG. Experimental results show that KREPE achieves state-of-the-art performance on standard HKG link prediction benchmarks while substantially outperforming large language model baselines in generating novel and correct hyper-relational facts.