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..
HyperMixup: Hypergraph-Augmented with Higher-order Information Mixup
Authors: Kaixuan Yao, Zhuo Li, Jianqing Liang, Jiye Liang, Ming Li, Feilong Cao
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we evaluate our proposed Hyper Mixup on two tasks: citation network classification and visual object recognition. We also compare the proposed method with graph convolutional networks and other state-of-the-art methods. Table 1: Summary of the citation classification datasets. Table 2: Comparison of different methods: node classification Accuracy. Figure 2: Test performance comparison for Hyper Mixup,GCN, HGNN, and HGNNp on Cora with different low label rates. |
| Researcher Affiliation | Academia | 1School of Computer and Information Technology, the Key Laboratory of Computational Intelligence and Chinese Information Processing of Ministry of Education, Shanxi University, Taiyuan, China 2Zhejiang Key Laboratory of Intelligent Education Technology and Application, Zhejiang Normal University, Jinhua, China 3School of Mathematics, Institute of Mathematics and Cross-disciplinary Science, Zhejiang Normal University, China |
| Pseudocode | No | The paper describes the methodology in narrative text and equations within Section 3. While it outlines a structured approach, it does not present any explicitly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | Yes | Question: Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [Yes] Justification: We have provided open access to the necessary data and code. |
| Open Datasets | Yes | 1) Citation Network Classification. Three benchmark datasets Cora, Pub Med, and Cite Seer [23] are adopted following the experimental protocol of HGNN [4]. ... 2) Visual Object Recognition. Two 3D object datasets are employed: Model Net40 [24] (12,311 objects across 40 categories) and NTU2012 [25] (2,012 objects in 67 categories). |
| Dataset Splits | Yes | Dataset Cora Pumbed Cite Seer Model Net40 NTU2012 Nodes 2708 19717 3327 12311 2012 Training node 140 60 120 9843 1639 Validation node 500 500 500 2468 373 Testing node 1000 1000 1000 40 67 Following the 80-20 train-test split convention, we extract multi-view features using MVCNN [26] and GVCNN [27]. |
| Hardware Specification | Yes | Experimental environment information is as follows: Intel(R) Xeon(R) Gold 6254 CPU @ 3.10GHz, 36 kernel, 512 G memory, NVIDIA RTX 3090 GPU. |
| Software Dependencies | No | The experimental setup follows the settings in HGNN[4].The following hyperparameters are set for all datasets: Adam optimizer with learning rate lr = 0.001. Layer number L = 2 with hidden dimension F = 16; In the reinforcement mixup module, we set p = 0.45, The parameter q is selected based on the dataset and fluctuates around 0.72, The parameter l is determined based on the selection of the dataset, resulting in a varying proportion of nearest neighbor samples. |
| Experiment Setup | Yes | The experimental setup follows the settings in HGNN[4].The following hyperparameters are set for all datasets: Adam optimizer with learning rate lr = 0.001. Layer number L = 2 with hidden dimension F = 16; In the reinforcement mixup module, we set p = 0.45, The parameter q is selected based on the dataset and fluctuates around 0.72, The parameter l is determined based on the selection of the dataset, resulting in a varying proportion of nearest neighbor samples. |