Do Temporal Knowledge Graph Embedding Models Learn or Memorize Shortcuts?
Abstract
Temporal Knowledge Graph Embedding models predict missing facts in temporal knowledge graphs. Previous work on static knowledge graph embedding models has revealed that KGE models utilize shortcuts in test set leakage to achieve high performance. In this work, we show that a similar test set leakage problem exists in widely used temporal knowledge graph datasets ICEWS14 and ICEWS05-15. We propose a naive rule-based model that can achieve start-of-the-art results on both datasets without a deep-learning process. Following this consideration, we construct two more challenging datasets for the evaluation of TKGEs.
Cite
Text
Pan et al. "Do Temporal Knowledge Graph Embedding Models Learn or Memorize Shortcuts?." NeurIPS 2023 Workshops: TGL, 2023.Markdown
[Pan et al. "Do Temporal Knowledge Graph Embedding Models Learn or Memorize Shortcuts?." NeurIPS 2023 Workshops: TGL, 2023.](https://mlanthology.org/neuripsw/2023/pan2023neuripsw-temporal/)BibTeX
@inproceedings{pan2023neuripsw-temporal,
title = {{Do Temporal Knowledge Graph Embedding Models Learn or Memorize Shortcuts?}},
author = {Pan, Jiaxin and Nayyeri, Mojtaba and Li, Yinan and Staab, Steffen},
booktitle = {NeurIPS 2023 Workshops: TGL},
year = {2023},
url = {https://mlanthology.org/neuripsw/2023/pan2023neuripsw-temporal/}
}