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..
Topology-aware Graph Diffusion Model with Persistent Homology
Authors: Joonhyuk Park, Donghyun Lee, Yujee Song, Guorong Wu, Won Hwa Kim
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on conventional graph benchmarks demonstrate the effectiveness of our approach indicating high generation performance across various metrics, while achieving closer alignment with the distribution of topological features observed in the original graphs. |
| Researcher Affiliation | Collaboration | 1POSTECH 2Samsung Eletronics 3UNC-Chapel Hill EMAIL |
| Pseudocode | Yes | Algorithm 1 Overall scheme of TAGG |
| Open Source Code | No | Furthermore, we will release the source code upon acceptance. |
| Open Datasets | Yes | We utilize open datasets (Community-small, Ego-small, ENZYMES and ADNI) and open-source TDA libraries (GUDHI) with detailed descriptions on the model architecture (Sec.4) and a minimum number of hyperparameters (Appendix A.3). |
| Dataset Splits | Yes | For consistency, we adhered to the train/test split reference from [23] and performed three replicate experiments to report averaged performance. |
| Hardware Specification | Yes | All experiments were conducted on a single Ge Force RTX 3090 with 24GB of GPU memory, with batch size 4. |
| Software Dependencies | No | We utilize open datasets (Community-small, Ego-small, ENZYMES and ADNI) and open-source TDA libraries (GUDHI) with detailed descriptions on the model architecture (Sec.4) and a minimum number of hyperparameters (Appendix A.3). |
| Experiment Setup | Yes | As explained in Sec. 4.3, the training objective of TAGG has two real valued hyperparameters α1 (0, 1] and α2 (0, 1], each used to control the cross-entropy loss of edges LE CE and the persistence diagram matching (PDM) loss LP DM, respectively. The hyperparameters α1 and α2 were chosen through a grid search of values in {1, 0.1, 0.01, 0.001, 0.0001} on each dataset, and the settings are shown in Tab. 6. Table 6: Hyperparameters of TAGG on different datasets Hyperparameter ADNI ENZYMES Community-small Ego-small α1 1 1 0.001 0.01 α2 0.001 0.0001 0.001 0.0001 |