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..
Revolutionizing Graph Aggregation: From Suppression to Amplification via BoostGCN
Authors: Jiaxin Wu, Chenglong Pang, Guangxiong Chen, Jie Zhao
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on four real-world datasets demonstrate that Boost GCN outperforms existing state-of-the-art GCN models in both performance and efficiency. 5 Experiments In our experiments, the following questions will be answered: RQ1: How does the performance of Boost GCN compare with state-of-the-art GCN models? RQ2: How efficient is Boost GCN? |
| Researcher Affiliation | Academia | Jiaxin Wu School of Management Guangdong University of Technology, China EMAIL Chenglong Pang School of Computer Science and Technology Donghua University, China EMAIL Guangxiong Chen School of Management Guangdong University of Technology, China EMAIL Jie Zhao School of Management Guangdong University of Technology, China EMAIL |
| Pseudocode | No | The paper describes the Boost GCN model using mathematical equations (Eq. 1-15) and conceptual figures (Figure 2 'The overview of Boost GCN'), but does not include any clearly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | No | Our code can be available on https://github.com/Clong Pang/Boost GCN. However, the NeurIPS checklist states: 'The reproducible settings involved in the experiment are all given in our paper, and we promise that our code will be sent out after the paper is accepted.', indicating that the code is not yet publicly released. |
| Open Datasets | Yes | To comprehensively demonstrate the effectiveness of Boost GCN, we evaluate our model on four distinct datasets, including Movie Lens-100k (denoted by 100k) [29], Movie Lens-1M (denoted by 1M) [29], Gowalla (denoted by Gowa.) [30] and Yelp2018 (denoted by Yelp) [11], as detailed in Table 2. The datasets utilized in this paper are all publicly available and can be directly downloaded from their respective sources. |
| Dataset Splits | Yes | Each dataset is divided into training (60%), validation (20%), and test sets (20%). |
| Hardware Specification | Yes | Experiments are tested on the same Intel(R) Xeon(R) Platinum 8255C CPU @2.50GHz machine with a Ge Force RTX 2080Ti GPU. |
| Software Dependencies | No | The paper mentions 'Adam [38]' for optimization and the 'Xavier method [39]' for initialization but does not provide specific version numbers for these or any other software components. |
| Experiment Setup | Yes | To ensure a fair comparison, we set the embedding size to 64, a learning rate of 10 3 with Adam [38] and initialize the embedding parameters with the Xavier method [39], which is the same as other baselines. To show the high generalization of Boost GCN on different datasets without parameter tuning, we set the fixed parameters: a batch size of 512, β = {e} and log = 1 in Eq.(12). Also, we provide guidance on λ for different types of datasets to optimize model performance. We employ λ {1e 6} for datasets of #Interactions 1 106, λ = {1e 4} for datasets of #Interactions 5 105 and λ = {1e 5} for others. |