What is YAGO3-TC? YAGO3-TC is a semi-complete KG created from a random subgraph of the test and validation data of YAGO3-10. As the test data, YAGO3-TC contains 28,364 triples from which 2,976 ones of them labeled as positive triples. Submission instructions You can find the validation and test data here. To get the ranking and evaluation metrics, after training the triple classifier, please submit the labels of test triples in here. If you want to contribute to the improvement of YAGO3-TC benchmark please please drop an email to pezeshkp@uci.edu.

Leaderboard Rank Model F1 1 Tucker 22.3 2 DistMult 20.4 3 RotatE 19.4

---

Citation:

@inproceedings{pezeshkpourrevisiting,
        title={Revisiting Evaluation of Knowledge Base Completion Models},
        author={Pezeshkpour, Pouya and Tian, Yifan and Singh, Sameer},
        booktitle={Automated Knowledge Base Construction (AKBC)},
        year={2020}
}