Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Knowledge Graph Completion by Intermediate Variables Regularization
Authors: Changyi Xiao, Yixin Cao
NeurIPS 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we conduct experiments to verify the effectiveness of our regularization technique as well as the reliability of our theoretical analysis. |
| Researcher Affiliation | Academia | Changyi Xiao, Yixin Cao School of Computer Science, Fudan University EMAIL, EMAIL |
| Pseudocode | Yes | Algorithm 1 A pseudocode for IVR |
| Open Source Code | Yes | The code is available at https://github.com/changyi7231/IVR. |
| Open Datasets | Yes | We evaluate the models on three KGC datasets, WN18RR [Dettmers et al., 2018], FB15k237 [Toutanova et al., 2015] and YAGO3-10 [Dettmers et al., 2018]. |
| Dataset Splits | Yes | We use the filtered MRR and Hits@N (H@N) [Bordes et al., 2013] as evaluation metrics and choose the hyper-parameters with the best filtered MRR on the validation set. |
| Hardware Specification | Yes | The time is the AMD Ryzen 7 4800U CPU running time on the test set. |
| Software Dependencies | No | We use Adagrad [Duchi et al., 2011] with learning rate 0.1 as the optimizer. While Adagrad is specified, no version numbers for this optimizer or other software libraries (e.g., Python, PyTorch/TensorFlow) are provided. |
| Experiment Setup | Yes | We use Adagrad [Duchi et al., 2011] with learning rate 0.1 as the optimizer. We set the batch size to 100 for WN18RR dataset and FB15k-237 dataset and 1000 for YAGO3-10 dataset. We train the models for 200 epochs. The settings for total embedding dimension D and number of parts P are shown in Table 5. The settings for power α and regularization coefficients λi are shown in Table 6. |