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..
Beyond Node-Centric Modeling: Sketching Signed Networks with Simplicial Complexes
Authors: Wei Wu, Xuan Tan, Yan Peng, Ling Chen, FangFang Li, Chuan Luo
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments show that Edge Sketch+ matches state-of-the-art accuracy while significantly reducing runtime, achieving speedups of up to 546.07 compared to GNN-based methods2. ...5 Experimental Results In this section, we conduct the extensive experiments to evaluate the performance of the proposed Edge Sketch+ method. ...Table 1: Link sign prediction performance results. ...5.2 Ablation Study ...5.3 Scalability ...5.4 Hyper-parameter Sensitivity |
| Researcher Affiliation | Academia | 1School of Computer Science and Engineering, Central South University 2Australian Artificial Intelligence Institute, University of Technology Sydney 3Xiangjiang Laboratory 4School of Software, Beihang University EMAIL EMAIL EMAIL EMAIL ...This work was supported by Open Project of Xiangjiang Laboratory (No.25XJ03020), National Natural Science Foundation of China (No. 62302528, 62522201, 62172449, 62202025, 72374070), Hunan Provincial Natural Science Foundation of China (2022JJ3021,2025JJ20071), Beijing Natural Science Foundation (No. L241050), Young Elite Scientist Sponsorship Program by CAST (No. YESS20230566), CCF-Huawei Populus Grove Fund CCF-Huawei (No. CCF-Huawei FM2024005), High Performance Computing Center of Central South University, and Fundamental Research Fund Project of Beihang University. |
| Pseudocode | Yes | Algorithm 1 The Edge Sketch+ Algorithm Input: 𝑔= (V, E+ E ), 𝐷, 𝑅, {h(𝑑,𝑟) 𝑘 }2,𝐷,𝑅 𝑘=0,𝑑=1,𝑟=1 Output: {x(𝑅) 𝑒𝑖}|E+ E | 𝑖=1 1: Extract all 0-, 1and 2-simplexes in 𝑔 2: Build the simplex-based adjacency matrices B1 and B2 3: Build the 1-dimensional Hodge Laplacian L1 4: for 𝑟= 1, . . . , 𝑅do 5: [x(𝑟 1) 𝑛1 ; ...; x(𝑟 1) 𝑛𝑁0 ] Conv B1 ([x(𝑟 1) 𝑛1 ; ...; x(𝑟 1) 𝑛𝑁0 ]) 6: [x(𝑟 1) 𝑡1 ; ...; x(𝑟 1) 𝑡𝑁2 ] Conv B2 ([x(𝑟 1) 𝑡1 ; ...; x(𝑟 1) 𝑡𝑁2 ]) 7: [x(𝑟 1) 𝑒1 ; ...; x(𝑟 1) 𝑒𝑁1 ] Conv L1 ([x(𝑟 1) 𝑒1 ; ...; x(𝑟 1) 𝑒𝑁1 ]) 8: for 𝑖= 1, . . . , 𝑁1 do 9: x(𝑟) 𝑒𝑖 AGG({h(𝑑,𝑟) 1 x(𝑟 1) 𝑒𝑖 }𝐷 𝑑=1, {h(𝑑,𝑟) 0 x(𝑟 1) 𝑛 |x(𝑟 1) 𝑛 is boundary adjacent on B1}𝐷 𝑑=1, {h(𝑑,𝑟) 2 x(𝑟 1) 𝑡 |x(𝑟 1) 𝑡 is co-boundary adjacent on B2}𝐷 𝑑=1, {h(𝑑,𝑟) 1 x(𝑟 1) 𝑒 |x(𝑟 1) 𝑒 is adjacent on L1}𝐷 𝑑=1) 10: x(𝑟) 𝑒𝑖 sgn(x(𝑟) 𝑒𝑖) 11: end for 12: end for |
| Open Source Code | Yes | 2We have released the source code and the datasets in https://github.com/AIand BD/Edge Sketchplus. ...We provide the code in the attachment. |
| Open Datasets | Yes | All dataset information is provided in Table 6 of Appendix E, where "BTC-𝛼" stands for Bitcoin-alpha, "BTC-O" for Bitcoin-OTC, "Slash." for Slashdot, and "Epin." for Epinions. ...We implement the link sign prediction task on the following signed networks from the SNAP group5. 5https://snap.stanford.edu/data/index.html |
| Dataset Splits | Yes | We report the mean of binary-f1, accuracy, auc, macro-f1 and end-to-end runtime in the 5-fold cross validation in Table 1. |
| Hardware Specification | Yes | All the experiments are conducted on Linux with 2.90GHz 128 Intel Xeon Platinum 8375C CPU, 1.5T RAM and NVIDIA A10 GPU (24GB RAM). |
| Software Dependencies | No | The paper does not explicitly state the specific version numbers for software dependencies used in their implementation (e.g., Python, PyTorch, or other libraries). |
| Experiment Setup | Yes | For details on baselines and experimental settings, please refer to Appendix F. ...Finally, we set 𝑅= 1 on Bitcoin-alpha and 𝑅= 4 on Bitcoin-OTC for MPSketch, which runs out of memory on Slashdot and Epinions; we set 𝑅= 1 on Bitcoin-alpha and 𝑅= 1 on Bitcoin-OTC, 𝑅= 1 on Slashdot and 𝑅= 1 on Epinions for Edge Sketch+. We set the embedding dimension 𝐷as 300 and the cutoff time of 24 hours. |