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..
Robust Explanations of Graph Neural Networks via Graph Curvatures
Authors: Yazheng Liu, Xi Zhang, Sihong Xie, Hui Xiong
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We empirically show that our method outperforms six base explanation methods in robustness across nine datasets spanning node classification, link prediction, and graph classification tasks, improving fidelity in 80% of the cases and achieving up to a 10% relative improvement in robust performance. The code is available at https://github.com/yazhengliu/Robust_explanation_curvature. ... Empirically, on nine datasets across node classification, link prediction, and graph classification tasks, we demonstrate that our method improves robustness in six state-of-the-art GNN explanation methods, enhancing fidelity in 80% of the cases and achieving up to a 10% relative improvement in robustness. |
| Researcher Affiliation | Academia | 1 The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China 2 The Beijing University of Posts and Telecommunications, Beijing, China |
| Pseudocode | No | The paper describes the proposed method and objective functions in text and mathematical formulations (e.g., Equation (7)) but does not provide a structured pseudocode block or algorithm section. |
| Open Source Code | Yes | The code is available at https://github.com/yazhengliu/Robust_explanation_curvature. |
| Open Datasets | Yes | We evaluate our method on three tasks across nine datasets: Cora, Citeseer, and Pub Med for node classification; BC-OTC, BC-Alpha, and UCI for link prediction; and MUTAG, PROTEINS, and IMDB-BINARY for graph classification. Details are provided in Appendix D.1. ... Node classification datasets: Cora, Citeseer, and Pub Med [1] are citation networks. ... Graph classification datasets: MUTAG [44] represents atom graph... |
| Dataset Splits | No | For each dataset, we train a GNN on the training set according to the task. Two GNN architectures are evaluated, with implementation details provided in Appendix D.2. We apply six classical explanation methods to compute edge importance scores F(u, v) and calculate Ricci curvature and effective resistance for all edges. To determine the optimal λ in Equation (7), we split the explanation targets into training and test subsets. |
| Hardware Specification | No | Our experiments were done on a CPU with a kernel size of 32GB. |
| Software Dependencies | No | The paper does not explicitly state specific software dependencies with version numbers (e.g., Python, PyTorch, or TensorFlow versions used for implementation). |
| Experiment Setup | Yes | During training, we set the learning rate to 0.01, the dropout rate to 0.5 and the hidden size to 16. The model is trained and then fixed during the explanation stages. |