DT4KGR: Decision transformer for fast and effective multi-hop reasoning over knowledge graphs

Yi Xia, Junyong Luo, Gang Zhou, Mingjing Lan, Xiaohui Chen, Jing Chen

Abstract: Highlights•To the best of our knowledge, our proposed DT4KGR constitutes the first sequence-to-sequence multi-hop KGR model.•The sequence-to-sequence generation capability of the Decision Transformer enables DT4KGR to coverage faster while being trained, and the genetated paths make the reasoning results more plausible and acceptable to humans.•Experimental results show that DT4KGR eventually conducts more efficient and effective multi-hop knowledge graph reasoning.