Research map / Federated learning and privacy / Personalized and heterogeneous FL
FedPerfix: Towards Partial Model Personalization of Vision Transformers in Federated Learning
Guangyu Sun, Matias Mendieta, Jun Luo, Shandong Wu, Chen Chen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add adapter-generated prefixes (FedPerfix) instead of vanilla attention personalization | ViT-Small PFL · CIFAR-100 | Top-1 accuracy: 44.63 -> 48.10 | Improved |
| generate prefixes with parallel adapter instead of zero-initialized prefixes (ablation) | FedPerfix · CIFAR-100 | Top-1 accuracy: 47.37 -> 48.10 | Improved |
| generate prefixes with parallel adapter instead of random-init prefixes (ablation) | FedPerfix · CIFAR-100 | Top-1 accuracy: 46.98 -> 48.10 | Improved |
| add personalized prompts on inputs (ablation) | ViT-Small PFL · CIFAR-100 | Top-1 accuracy: 44.19 vs 44.42 (head local) | No clear effect |
| use MLP adapters instead of prefixes (ablation) | FedPerfix · CIFAR-100 | Top-1 accuracy: 48.10 -> 47.99 | No clear effect |
| keep self-attention layers local alongside head (ablation) | ViT-Small FedAvg · CIFAR-100 | Top-1 accuracy: 23.29 -> 42.95 | Improved |
| keep LayerNorm local (FedBN-style) (ablation) | ViT-Small FedAvg · CIFAR-100 | Top-1 accuracy: 23.29 -> 23.94 | No clear effect |
| replace ResNet50 with ViT-Small backbone (ablation) | FedRep and other PFL methods · CIFAR-100 | Top-1 accuracy: FedRep 39.52 -> 44.42 | Improved |
| increase model size from ViT-Small to ViT-Base (ablation) | FedPerfix · CIFAR-100 | Top-1 accuracy: 48.10 -> 48.40 | Improved |
| lower client participation rate to 6.25% (ablation) | FedPerfix · CIFAR-100 (N=64) | Top-1 accuracy: 48.10 -> 43.80 | Hurt |
| Per-FedAVG meta-learning on ViT | FedAvg · Office-Home | Top-1 accuracy: 21.47 -> 17.09 | Hurt |
Later papers that compared with this method
1 later accepted paper reports FedPerfix as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- FedNASP: Federated Vision-Language Navigation with Adaptive Step-wise Personalization (ECCV 2026 · better on 3 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Personalized Federated Learning via Feature Distribution Adaptation (NeurIPS 2024 · similarity 0.90)
- Unlocking the Potential of Prompt-Tuning in Bridging Generalized and Personalized Federated Learning (CVPR 2024 · similarity 0.88)
- Flexible Sharpness-Aware Personalized Federated Learning (AAAI 2025 · similarity 0.87)
- Tackling Feature-Classifier Mismatch in Federated Learning via Prompt-Driven Feature Transformation (NeurIPS 2025 · similarity 0.87)
- Harmonizing Generalization and Personalization in Federated Prompt Learning (ICML 2024 · similarity 0.87)
- BLEND: Balancing Personalization vs. Generalization in Federated Vision–Language Models (NeurIPS 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
