Prototype-based federated learning has emerged as a promising approach that shares lightweight prototypes to transfer knowledge among clients with data heterogeneity in a model-agnostic manner. However, existing methods often collect prototypes directly from local models, which inevitably introduce inconsistencies into representation learning due to the biased data distributions and differing model architectures among clients. In this paper, we identify that both statistical and model heterogeneity create a vicious cycle of representation inconsistency, classifier divergence, and skewed prototype alignment, which negatively impacts the performance of clients. To break the vicious cycle, we propose a novel framework named Federated Learning via Semantic Anchors (FedSA) to decouple the generation of prototypes from local representation learning. We introduce a novel perspective that uses simple yet effective semantic anchors serving as prototypes to guide local models in learning consistent representations. By incorporating semantic anchors, we further propose anchor-based regularization with margin-enhanced contrastive learning and anchor-based classifier calibration to correct feature extractors and calibrate classifiers across clients, achieving intra-class compactness and inter-class separability of prototypes while ensuring consistent decision boundaries. We then update the semantic anchors with these consistent and discriminative prototypes, which iteratively encourage clients to collaboratively learn a unified data representation with robust generalization. Extensive experiments under both statistical and model heterogeneity settings show that FedSA significantly outperforms existing prototype-based FL methods on various classification tasks.

Although the fine-tuning-based deep transfer learning method performs well in adapting a pre-trained model to downstream tasks, it struggles in non-stationary environments where the data distribution changes dynamically over time. In such scenarios, the trained model rapidly becomes obsolete. Directly fine-tuning it for new environments will inevitably lead to negative transfer, degrading the model's generalization performance. Inspired by neuroscience, some existing studies have proposed a so-called “unlearning-and-relearning” training paradigm to alleviate this issue. However, existing methods face limitations; for example, they may destroy general features in shallow layers and lack adaptability when performing deep-layer unlearning. This paper proposes a Layer-wise Unlearning (LwU) method. Our approach first identifies the layers requiring unlearning by estimating the transferability of each layer in the trained model. Then, within the selected layers, it preserves high-sensitivity parameters and re-initializes low-sensitivity ones based on a parameter sensitivity analysis to achieve unlearning. Extensive experiments on synthetic, real-world, and image datasets demonstrate that LwU effectively overcomes negative transfer and substantially improves the model's generalization performance in new environments. The code of the proposed method and data are available at https://github.com/mlmmwym/LwU.

As AI evolves, collaboration among heterogeneous models helps overcome data scarcity by enabling knowledge transfer across institutions and devices.Traditional Federated Learning (FL) only supports homogeneous models, limiting collaboration among clients with heterogeneous model architectures.To address this, Heterogeneous Federated Learning (HtFL) methods are developed to enable collaboration across diverse heterogeneous models while tackling the data heterogeneity issue at the same time.However, a comprehensive benchmark for standardized evaluation and analysis of the rapidly growing HtFL methods is lacking.Firstly, the highly varied datasets, model heterogeneity scenarios, and different method implementations become hurdles to making easy and fair comparisons among HtFL methods.Secondly, the effectiveness and robustness of HtFL methods are under-explored in various scenarios, such as the medical domain and sensor signal modality.To fill this gap, we introduce the first Heterogeneous Federated Learning Library (HtFLlib), an easy-to-use and extensible framework that integrates multiple datasets and model heterogeneity scenarios, offering a robust benchmark for research and practical applications.Specifically, HtFLlib integrates (1) 12 datasets spanning various domains, modalities, and data heterogeneity scenarios; (2) 40 model architectures, ranging from small to large, across three modalities;(3) a modularized and easy-to-extend HtFL codebase with implementations of 10 representative HtFL methods; and (4) systematic evaluations in terms of accuracy, convergence, computation costs, and communication costs.We emphasize the advantages and potential of state-of-the-art HtFL methods and hope that HtFLlib will catalyze advancing HtFL research and enable its broader applications.The code is released at https://github.com/TsingZ0/HtFLlib.