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..
Tackling Feature-Classifier Mismatch in Federated Learning via Prompt-Driven Feature Transformation
Authors: Xinghao Wu, Xuefeng Liu, Jianwei Niu, Guogang Zhu, Mingjia Shi, Shaojie Tang, Jing Yuan
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments show that Fed PFT outperforms state-of-the-art methods by up to 5.07%, with further gains of up to 7.08% when collaborative contrastive learning is incorporated. The code is available at https://github.com/Xinghao Wu/Fed PFT. |
| Researcher Affiliation | Academia | 1State Key Laboratory of Virtual Reality Technology and Systems, School of Computer Science and Engineering, Beihang University, Beijing, China; 2Hangzhou Innovation Institute of Beihang University, Zhejiang Key Laboratory of Industrial Big Data and Robot Intelligent Systems, Hangzhou, China 3 Center for AI Business Innovation, Department of Management Science and Systems, University at Buffalo, Buffalo, New York, USA 4 University of North Texas, Denton, Texas, USA 5 Zhongguancun Laboratory, Beijing, China 6 Sichuan University, Sichuan, China |
| Pseudocode | Yes | The pseudo-code of Fed PFT is summarized in Algorithm 1. |
| Open Source Code | Yes | The code is available at https://github.com/Xinghao Wu/Fed PFT. |
| Open Datasets | Yes | Specifically, we employ three datasets: CIFAR-10 [14], CIFAR-100 [13], and Tiny Image Net [15]. Additionally, in Section 4.6. we also verify Fed PFT in the feature shift non-IID scenario with PACS [17] and Domain Net [38] datasets. |
| Dataset Splits | Yes | In the label shift scenario, each client is assigned 500 training samples. For CIFAR-10 and CIFAR100 datasets, each client has 100 test samples; for the Tiny Image Net dataset, each client has 200 test samples. Both training and test data have the same label distribution. In each setting, we employ three datasets: CIFAR-10 [14], CIFAR-100 [13], and Tiny Image Net [15]. |
| Hardware Specification | Yes | Experiments are implemented using Py Torch and conducted on 4x NVIDIA RTX 2080 GPUs. |
| Software Dependencies | No | Experiments are implemented using Py Torch and conducted on 4x NVIDIA RTX 2080 GPUs. |
| Experiment Setup | Yes | For the general hyperparameters of FL, we set the number of clients N = 40, batch size B = 100, and local update rounds R = 5. The total rounds T are set to 1000 to ensure all methods reach full convergence. We select the highest average accuracy achieved by all clients across all rounds as the result. Each experiment is repeated with three random seeds, and the mean and standard deviation are reported. We employ the Res Net [9] model architecture, specifically Res Net-8 for CIFAR-10 and Res Net-10 for CIFAR-100 and Tiny Image Net. Please refer to the Appendix F for more details. |