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..
Rising from Ashes: Generalized Federated Learning via Dynamic Parameter Reset
Authors: Jiahao Wu, Ming Hu, Yanxin Yang, Xiaofei Xie, ZeKai Chen, Chenyu Song, Mingsong Chen
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on various well-known datasets demonstrate that compared to SOTA FL methods, Fed Phoenix can achieve up to 20.73% higher accuracy. The implementation is publicly available at https://github.com/Uni String/Fed Phoenix. |
| Researcher Affiliation | Academia | 1Mo E Eng. Research Center of SW/HW Co-Design Tech. and App., East China Normal University 2School of Computing and Information Systems, Singapore Management University |
| Pseudocode | Yes | Algorithm 1 Fed Phoenix Framework Algorithm 2 Model Reset |
| Open Source Code | Yes | The implementation is publicly available at https://github.com/Uni String/Fed Phoenix. |
| Open Datasets | Yes | Experimental results on various well-known datasets demonstrate that compared to SOTA FL methods, Fed Phoenix can achieve up to 20.73% higher accuracy. The implementation is publicly available at https://github.com/Uni String/Fed Phoenix. |
| Dataset Splits | No | The paper mentions partitioning datasets via IID sampling and non-IID Dirichlet Distribution, and discusses training and testing datasets, but does not provide specific percentages or counts for train/test/validation splits. |
| Hardware Specification | Yes | All the experimental results were obtained from an Ubuntu workstation with an Intel i9 CPU, 256GB of memory, and an NVIDIA RTX 4090 GPU. |
| Software Dependencies | No | The paper mentions using SGD optimizer and specific models (Res Net-18, Mobile Net-V1, VGG-16), but does not provide specific version numbers for software libraries or frameworks (e.g., Python, PyTorch, TensorFlow, CUDA). |
| Experiment Setup | Yes | For all experiments, we employed the SGD optimizer with a learning rate of 0.01 and a momentum of 0.5. Each FL training round utilizes a batch size of 50 and 5 local training epochs. For Fed Phoenix, we set the reset rate to θ = 1/32 for both IID and non-IID scenarios with β = 0.6, and set θ = 1/64 for non-IID scenarios with β = 0.3. |