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 Straggler-Resilient Split Federated Learning: An Unbalanced Update Approach
Authors: Dandan Liang, Jianing Zhang, Evan Chen, Zhe Li, Rui Li, Haibo Yang
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments demonstrate that MU-Split Fed consistently outperforms baseline methods with the presence of stragglers and effectively mitigates their impact through adaptive tuning of τ. |
| Researcher Affiliation | Academia | Dandan Liang Rochester Institute of Technology Rochester, New York EMAIL; Jianing Zhang Purdue University West Lafayette, Indiana EMAIL; Evan Chen Purdue University West Lafayette, Indiana EMAIL; Zhe Li Rochester Institute of Technology Rochester, New York EMAIL; Rui Li Rochester Institute of Technology Rochester, New York EMAIL; Haibo Yang Rochester Institute of Technology Rochester, New York EMAIL |
| Pseudocode | Yes | Algorithm 1: MU-Split Fed |
| Open Source Code | Yes | The code for this project is available at https://github.com/Johnny-Zip/MU-Split Fed. |
| Open Datasets | Yes | To evaluate the effectiveness of MU-Split Fed, we conduct experiments on four image classification benchmarks: Fashion-MNIST [31], CINIC-10 [32], CIFAR-10, and CIFAR-100 [33]. We evaluate its performance on the SST-2 dataset [35]. |
| Dataset Splits | No | In our experiment, we train 10 clients in total with 50% partial partitioning for each global aggregation. |
| Hardware Specification | Yes | All experiments are carried out on a node with 3 NVIDIA A100 40GB GPUs. |
| Software Dependencies | No | The paper does not provide specific version numbers for software dependencies. |
| Experiment Setup | Yes | Table 6: Hyperparameters PARAMETER VALUE EXPLANATION ηg 0.3 Global aggregation learning rate ηs 0.01 Server learning rate ηc 0.005 Client learning rate λ 0.005 Scale of perturbation for ZOO B 32 Batch size |