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..
Future Link Prediction Without Memory or Aggregation
Authors: Lu Yi, Runlin Lei, Fengran Mo, Yanping Zheng, Zhewei Wei, Yuhang Ye
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on diverse datasets demonstrate that CRAFT consistently achieves superior performance with high efficiency, making it well-suited for large-scale real-world applications. |
| Researcher Affiliation | Collaboration | 1Renmin University of China, 2Université de Montréal EMAIL EMAIL 3Huawei Poisson Lab, Huawei Technology Ltd. EMAIL |
| Pseudocode | No | The paper describes the model architecture and mathematical formulations in Section 4 'Proposed Methods' and its sub-sections, but does not include any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | The implementation of CRAFT and all commands to reproduce the experimental results are publicly available at https: //github.com/luyi256/CRAFT. |
| Open Datasets | Yes | All datasets are publicly available. |
| Dataset Splits | Yes | We follow the evaluation protocol of TGB-Seq and TGB benchmarks for all datasets. We use MRR as the evaluation metric, follow the original split of datasets, or chronologically split the datasets into training, validation, and test sets with 75%, 15%, and 15% of the edges, respectively. |
| Hardware Specification | Yes | Missing entries indicate cases where the model failed due to either out-of-time (unable to complete a single training epoch within 24 hours) or out-of-memory on a 32GB GPU. |
| Software Dependencies | No | The implementation of CRAFT and baselines are based on Dy GLib [46]. |
| Experiment Setup | Yes | We follow the configurations in [44, 13, 46] to use the same batch size (200 or 400, accordingly) and learning rate (1e-4) across all methods. The final hyperparameters for CRAFT, along with the batch size and embedding size per dataset, are summarized in Table 8. |