Boosting Knowledge Graph Foundation Models via Enhanced Negative Sampling
The paper discusses an innovative approach to improve Knowledge Graph Foundation Models (KGFMs) through enhanced negative sampling. The proposed method, KMAS, constructs hard negative triples to provide better training supervision for KGFMs. Experimental results indicate that this method can significantly enhance the performance of KGFMs without requiring excessive computational resources.
- ▪Knowledge graphs are essential for tasks like question answering and recommender systems but often lack completeness.
- ▪The KMAS method dynamically adjusts the ratio of hard negative triples during training to improve model performance.
- ▪Extensive experiments on 44 datasets show that KMAS enhances state-of-the-art KGFMs efficiently.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.27023 (cs) [Submitted on 26 May 2026] Title:Boosting Knowledge Graph Foundation Models via Enhanced Negative Sampling Authors:Yinan Liu, Wenjin Xu, Zhiyuan Zha, Xiaochun Yang, Bin Wang View a PDF of the paper titled Boosting Knowledge Graph Foundation Models via Enhanced Negative Sampling, by Yinan Liu and 4 other authors View PDF HTML (experimental) 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.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
