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..

Towards Unsupervised Open-Set Graph Domain Adaptation via Dual Reprogramming

Authors: Zhen Zhang, Bingsheng He

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 4 Experiments 4.1 Experimental Settings 4.2 Results and Analyses 4.3 Ablation Studies
Researcher Affiliation Academia Zhen Zhang1,2 1Nanjing University 2National University of Singapore EMAIL Bingsheng He National University of Singapore Singapore EMAIL
Pseudocode Yes Algorithm 1 Graph RTA s Training Strategy
Open Source Code Yes Our source codes and datasets are publicly available at https://github.com/cszhangzhen/Graph RTA.
Open Datasets Yes Our source codes and datasets are publicly available at https://github.com/cszhangzhen/Graph RTA. To thoroughly assess the performance of our proposed Graph RTA, we conduct experiments using three categories of publicly available datasets. An overview of these dataset characteristics is provided in Table 1
Dataset Splits Yes Among them, 70% of the labeled source nodes are utilized for training, 10% are set aside for validation, and the remaining 20% serve as a sanity check. The ๏ฌnal evaluation is conducted on the target nodes.
Hardware Specification Yes Our experiments are conducted on a Linux server with 2 AMD EPYC 7543 CPU@2.80GHz, 512G RAM and one NVIDIA A100-SXM4-80GB GPU.
Software Dependencies Yes The proposed model is implemented with Pytorch 1.13.1 in Python 3.8 using Pytorch Geometric 2.4.0.
Experiment Setup Yes Hyperparameters for learning rate, weight decay and ฮป are searched within the ranges of [0.1, 0.01, 0.001, 1e 4, 1e 5], and the sparse constraint ฯ is explored within the interval [0, 1]. The experiments are repeated ๏ฌve times, and performance metrics are reported as the mean along with standard deviations for both accuracy and H-score [10].