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..
Coloring Learning for Heterophilic Graph Representation
Authors: Miaomiao Huang, Yuhai Zhao, Daniel Zhengkui Wang, Fenglong Ma, Yejiang Wang, Meixia Wang, Xingwei Wang
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | This section empirically evaluates the proposed Co Rep method on 14 benchmark datasets and analyzes its behavior on graphs to gain further insights. More results can be found in the Appendix F. |
| Researcher Affiliation | Academia | Miaomiao Huang1,2, Yuhai Zhao1,2, , Zhengkui Wang3, Fenglong Ma4, Yejiang Wang1,2, Meixia Wang1,2, Xingwei Wang1 1 School of Computer Science and Engineering, Northeastern University, China 2 Key Laboratory of Intelligent Computing in Medical Image of Ministry of Education, Northeastern University, China 3 Info Comm Technology Cluster, Singapore Institute of Technology, Singapore 4 College of Information Sciences and Technology, Pennsylvania State University, United States EMAIL, EMAIL, EMAIL, EMAIL |
| Pseudocode | Yes | We present the overall algorithm in Appendix B. |
| Open Source Code | Yes | Answer: [Yes] Justification: We provide the source code and data as supplementary material in the submission. |
| Open Datasets | Yes | Datesets. To assess the quality of the learned representations, we employ transductive node classification as the downstream task. Our experiments are conducted on 14 widely used benchmark datasets, consisting of 8 homophilic graph datasets (i.e., Cora, Cite Seer, Pub Med, Wiki-CS, Amazon Computers, Amazon Photo, Co Author CS, and Co Author Physics) [38, 29, 39] and 6 heterophilic graph datasets (i.e., Chameleon, Squirrel, Actor, Cornell, Texas, and Wisconsin) [31]. The statistics of all datasets are summarized in Appendix D. |
| Dataset Splits | Yes | For datasets, we adopt the standard dataset splits used in previous studies, i.e., public splits [52, 22, 31] or commonly used splits [60, 27]. |
| Hardware Specification | No | The paper states: "We provide the experimental platform in the main paper and appendix, including the settings of hardware, hyper-parameters, and so on, as well as providing the time complexity of the model." However, specific details such as GPU/CPU models, memory, or specific computing environments are not explicitly provided within the given text. |
| Software Dependencies | No | All methods were implemented in Py Torch with the Adam Optimizer. We run 10 times of experiments and report the average test accuracy with standard deviation. For fair comparison, the parameters of all baselines are tuned according to the parameter ranges reported by the authors. Specific hyperparameter settings and more implementation details are in Appendix E. |
| Experiment Setup | Yes | All methods were implemented in Py Torch with the Adam Optimizer. We run 10 times of experiments and report the average test accuracy with standard deviation. For fair comparison, the parameters of all baselines are tuned according to the parameter ranges reported by the authors. Specific hyperparameter settings and more implementation details are in Appendix E. |