Boosting Knowledge Graph Foundation Models via Enhanced Negative Sampling

2026-05-26

AI Summary

Current knowledge graph foundation models suffer from weak supervision and limited generalization in zero-shot link prediction due to their reliance on low-quality randomly sampled negative triples during training. To address this, this work proposes KMAS, an adaptive negative sampling method that dynamically constructs hard negatives using relation embeddings generated by a relation encoder. Furthermore, it introduces a linear scheduling strategy—first increasing and then decreasing the proportion of hard negatives—to align with the evolving learning capacity of the model throughout training. The approach incurs negligible computational overhead and consistently enhances the zero-shot link prediction performance of multiple state-of-the-art models across 44 datasets, demonstrating the effectiveness of relation embedding–guided hard negative construction and dynamic scheduling.

Abstract

Knowledge graphs (KGs) have become the core backbone of numerous downstream tasks such as question answering and recommender systems. However, despite all this, KGs are often very incomplete. To perform zero-shot knowledge graph completion in unseen KGs, which have different relational vocabularies from those used for pre-training, KG foundation models (KGFMs) receive a wide range of attention. Existing KGFMs often perform training using random negative triples, which are constructed by replacing the head or tail entity of a positive triple with a random entity. However, these negative triples are often constructed with limited quality, providing weak supervision for KGFM training. In this paper, we propose a simple yet effective adaptive negative sampling approach, KMAS, to enhance existing KGFMs. KMAS constructs hard negative triples through the updated relation embeddings generated from the existing KGFM's relation encoder. To further adaptively align with the evolving capability of the KGFM during the training process, KMAS adjusts the ratio of hard negative triples dynamically throughout the whole training process: after a warmup phrase, it increases the ratio linearly and then decreases linearly. Extensive experiments are conducted over 44 data sets. Experimental results demonstrate that our proposed negative sampling method can enhance many SOTA KGFMs without requiring excessive additional time or memory consumption.

Problem

• Knowledge Graph Foundation Models
• Negative Sampling
• Knowledge Graph Completion
• Zero-shot Learning
• Hard Negatives

Innovation

• adaptive negative sampling
• knowledge graph foundation models
• hard negative triples
• relation embeddings
• zero-shot knowledge graph completion